## On MicroSD Problems

The microSD ware for January 2010 was not an incidental post. It is actually snapshot of a much longer forensic investigation to find the ground truth behind some irregular Kingston memory cards.

It all started back in December of 2009, when chumby was in the midst of production for the chumby One. A call came in from the floor noting that SMT yield had dropped dramatically on one lot, so I drove over to the building to have a look (this is the advantage of being in China during production — you can fix problems like this within the hour, before they become really serious issues). After poking and prodding a bit, I realized that all the units failing had Kingston microSD cards from a particular lot code. I had the factory pull the entire lot of microSD cards from the line and rework all the units that had these cards loaded. Sure enough, after subtracting these cards from the line, yield was back to normal again.

Normally, the story would end there; you’d RMA the material, get an exchange for the lot, and move on. Except there were a couple of problems. First, Kingston wouldn’t take the cards back because we had programmed them. Second, there was a lot of them — about a thousand all together, and chumby was already deeply back-ordered. Also, memory cards aren’t cheap; the spot price on this type of memory card is around 4-5, so it’s a few thousand dollars in scrap if we can’t get them exchanged … and neither chumby nor the CM is large enough to sneeze at a few kilobucks. So I kicked into forensic mode. The first thing that raised my suspicions is the external markings on the irregular Kingston cards. On the left is a sample of the irregular card. On the right is a sample of a normal card. I’ve put red arrows on the details that called the most attention to me at first. The most blatantly strange issue is that the card on the left has its lot code silkscreened using the same stencil as the main logo. Silkscreening a lot code on isn’t that unusual, but typically the silk does not share the same stencil as the logo, so you’ll see some small variance in the coloration, font, or alignment of the lot code from the rest of the text. In fact, across the entire batch of irregular cards, they shared the exact same lot code (N0214-001.A00LF) (typically the lot code will vary every couple hundred cards at least). This is in contrast to the card on the right, which is laser-marked, and has a lot code that varied with every tray of 96 units. The second strange issue, perhaps more subtle and perhaps not damning, is the irregularity in the “D” of the microSD logo. Typically, brand name vendors like Kingston would be very picky about the accuracy of their logos. The broken D is something found on SanDisk cards, but Kingston cards found in US retail almost universally use a solid D. It turns out the weirdness in the external markings is just the start of it. When we read out the electronic card ID data on the two cards (available through /sys entries in linux), this is what we found: First, the date code on the irregular card is uninitialized. Dates are counted as the offset from 00/2000 in the CID field, so a value of 00/2000 means they didn’t bother to assign a date (for what it’s worth, in the year 2000, 2GB microSD cards also didn’t exist). Also, the serial number is very low — 0x960 is decimal 2,400. Other cards in the irregular batch also had similarly very low serial numbers, in the hundreds to thousands range. The chance of me “just happening” to get the very first microSD cards out of a factory is pretty remote. The serial number of the normal card, for example, is 0x9C62CAE6, or decimal 2,623,720,166 — a much more feasible serial number for a popular product like a microSD card. Very low serial numbers, like very low MAC ID addresses, are a hallmark of the “ghost shift”, i.e. the shift that happens very late at night when a rouge worker enters the factory and runs the production machine off the books. Significantly, ghost shifts are often run using marginal material that would normally be disposed of but were intercepted on the way to the grinder. As a result, the markings and characteristics of the material often look absolutely authentic, because the ghost material is a product of the same line as genuine material. Furthermore, the manufacturer’s ID is 0x41 (ASCII ‘A’), which I don’t recognize (supposedly the SD group assigns all the MIDs but I don’t see a public list of them anywhere). The OEMID is also 0x3432, which is suspiciously ASCII ’42’ (one more than the hex value for the manufacturer ID). These hex/ascii confusions are possible signs that someone who didn’t appreciate the meaning of these fields was running a ghost shift making these cards. Armed with this evidence, we confronted Kingston — both the distributor in China as well as the US sales rep. First, we wanted to know if these were real cards, and second, if they were real cards, why were the serialization codes irregular? After some time, the Kingston guys came back to us and swore these cards were authentic, not fakes, but at least they reversed their position on not offering an exchange on the cards — they took back the programmed cards and exchanged them for new ones, no further questions asked. However, they never answered as to why their card ID numbers were irregular. While I know chumby is a small fry customer compared to the Nokias of the world, I think it’s still important that they answer basic questions about their quality control process even to the small fry. I had an issue once with an old version of a Quintic part being accidentally shipped to me, and once I could prove the issue to them, I received world-class customer service from Quintic, a full explanation, and an immediate and full exchange of the parts at their cost. That was exemplary service, and I commend and strongly recommend Quintic for it. Kingston, on the other hand, did not set an example to follow. Normally, at this point, I would simply disqualify Kingston as a vendor, but I’m more persistent than that. It’s disconcerting that a high-profile, established brand would stand behind such irregular components. Who is to say SanDisk or Samsung wouldn’t do the same? Price erosion has been brutal on all the FLASH vendors, and as small fry I might be repeatedly taken advantage of as a sink for marginal material to improve the FLASH vendor’s bottom lines. Given the relatively high cost of these components, I needed to develop some simple guidelines for IQC (incoming quality control) inspection to accept or reject shipments from memory vendors, so I decided to do more digging to try and find ground truth. The first thing I had to do was collect a lot of samples. The key is to attempt to collect both regular and irregular cards in the wild, so I went to the SEG / Hua Qian Bei district and wandered around the gray markets there. I bought about ten memory cards total from small vendors, at prices varying from 30-50 RMB (4.40 – $7.30), most of them priced toward 30 RMB. The process of shopping for irregular cards itself was interesting. In talking to a couple dozen vendors, you learn a few things. First, Kingston as a brand is weak in China for microSD cards. Sandisk has done a lot more marketing in the microSD space, and as a result, it’s much easier to find Sandisk cards on the open market. The quality of the grey-market Sandisk cards are also typically more consistent. Second, the small vendors are entirely brazen about selling you well-crafted fakes. Typically, the bare cards are just sitting loose in trays in the display case; once you agree on the price and commit to buying the card, the vendor will toss the loose card into a “real” Kingston retail package, and then miraculously pull out a certificate, complete with hologram, serial numbers, and a kingston.com URL you can visit to validate your purchase, and slap it on the back of the retail package right in front of your eyes. Hey, it’s just like new! … I suppose the typical buyer in those markets is not an end user, but someone who is looking to make a quick buck reselling these cards at a hefty markup in a more reputable retail outlet. One vendor in particular interested me; it was literally a mom, pop and one young child sitting in a small stall of the mobile phone market, and they were busily slapping dozens of non-Kingston marked cards into Kingston retail packaging. They had no desire to sell to me, but I was persistent; this card interested me in particular because it also had the broken “D” logo but no Kingston marking. Above is a scan of the card and the package it came in (a larger image of the card can be seen below; it is “Sample #4”). After collecting all the samples, I read out their card ID information, and then digested their packages with nitric acid. Below is the line-up of the cards I digested. Yes, my digestion technique is pretty crude. Actually, most of the damage to the card came from the cleaning process — I was using a Q-tip with acetone to remove the dissolved encapsulant and I had to get a little rough, which doesn’t do any favors for the bond wires. But…good enough for my purposes. Click on the image above for a full-sized version. Some notes on the cards above: • Sample 1: This is the original irregular card that got me started on this whole arc. It was purchased through a sanctioned Kingston distributor in China, and to the best of my knowledge, none were shipped to end customers of chumby. MID = 0x000041, OEMID = 0x3432, serial = 0x960, name = SD2GB. • Sample 2: This is a normal card that I also purchased from the same sanctioned Kingston distributor in China, and is typical of those actually shipped in the first lot of chumby Ones MID = 0x000002, OEMID = 0x544D, serial = 0x9C62CAE6, name = SA02G • Sample 3: This is a Kingston card purchased through a major US retail chain. Note how the MID and OEMID are identical to sample 2, but not sample 1. MID = 0x000002, OEMID = 0x544D, serial = 0xA6EDFA97, name = SD02G • Sample 4: This is the aforementioned non-Kingston branded card that I spotted being slapped into Kingston-marked packaging, bought on the open market in Shenzhen. Note the low serial number. MID = 0x000012, OEMID = 0x3456, serial = 0x253, name = MS • Sample 5: This is a device bought from a more established retailer in the Shenzhen market, but still questionable. I bought it because it had the XXX.A00LF marking, like my original irregular card. MID = 0x000027, OEMID = 0x5048, serial = 0x7CA01E9C, name = SD2GB • Sample 6: This is a SanDisk card bought on the open market from a sketchy shop run by a sassy chain-smoking girl who wouldn’t stop texting on her mobile. I actually acquired three total SanDisk cards from different sketchy sources but all of them checked out with the same CID info, so I only opened one of them. Interestingly, one SanDisk card turned out to be used and only quick-formatted. With the help of some recovery software, I found DLLs, WAV’s, maps, and verisign certificates belonging to Navione’s Careland GPS inside the drive. A project for another day will be acquiring lots of refurb microSD cards and collecting interesting data off of them. MID = 0x000003, OEMID = 0x5344, serial = 0x114E933D, name = SU02G • Sample 7: This is a Samsung card that we bought from a Samsung wholesale distributor. I didn’t scan this one before digesting it, so the image of it is missing but the card actually has no markings on the outside — it’s a total blank card with just a laser mark on the back. From appearances alone, it would look to be the sketchiest of the bunch, but in reality it’s one of the best built. Goes to show you can’t judge a book by its cover. MID = 0x00001B, OEMID = 0x534D, serial = 0xB1FE8A54, name = 00000 • That’s a lot of data for a blog post, but I figured more details are better for sharing, since I could find no central database for this kind of information on the web. Here are the most interesting “high level” results from my survey: • The “normal” Kingston cards (samples #2 and #3) were all direct Toshiba OEM cards (MID = 0x000002, OEMID = 0x544D (ASCII ‘TM’, presumably for Toshiba Memory)). These cards employ Toshiba controllers and Toshiba memory chips, and seem to be of good quality, and thankfully the only ones that were sent on to chumby customers. • The irregular card (sample #1) uses the same controller chip as the outright fake (sample #4) that was bought in the SZ market. Both the irregular Kingston and the fake Kingston had low serial numbers and whacky ID information. Some of these cards experience some difficulty in normal operation. I still hesitate to call Kingston’s irregular card a fake — that’s a very strong accusation to make — but its construction is similar to another card of clearly questionable quality, which leads me to question Kingston’s judgment in picking authorized manufacturing partners. • The irregular card is the only card in the group that does not use a stacked CSP construction. Instead, it uses side-by-side bonding. • The only two memory chip foundries in this sample set were Toshiba/Sandisk and Samsung. Note that Sandisk and Toshiba co-own the fab that makes their memory chips. • Samsung’s NAND die — the most expensive part of a microSD card — is about 17% larger than Toshiba/Sandisk. This means that Samsung microSD cards should naturally carry a slightly higher price than Toshiba/Sandisk cards. However, Samsung does get to offset that against the ability to diversify the same die from microSD packages into street-packaged TSOP devices, and they also don’t have a middleman like Kingston to eat away at margins. • Significantly, Kingston is revealed as simply a vendor that re-marks other people’s chips in its own packaging [clarification]. Every Kingston card surprisingly had a Sandisk/Toshiba memory chip inside, and the only variance or “value add” that could be found is in the selection of the controller chip. Oddly enough, of all the vendors, Kingston quoted with the best lead times and pricing — better than SanDisk or Samsung, despite the competition making all their own silicon and thereby having a lower inherent cost structure. This tells me that Kingston must be crushed when it comes to margin, which may explain why irregular cards are finding their way into their supply chain. Kingston is also probably more willing to talk to smaller accounts like me because as a channel brand they can’t compete against OEMs like Sandisk or Samsung for the biggest contracts from the likes of Nokia or RIMM. Effectively, Kingston is just a channel trader and is probably seen by SanDisk/Toshiba as a demand buffer for their production output. I also wouldn’t be surprised if SanDisk/Toshiba was selling Kingston “A-” grade parts, i.e., parts with slightly more defective sectors, but otherwise perfectly serviceable. As a result, Kingston plays a significant and important role in stabilizing microSD card prices and improving fab margins, but at some risk to their own brand image. Overall, the MicroSD card market is a fascinating one, a discussion perhaps worth a blog post on its own. I’d like to point out to casual readers that the spot price of MicroSD cards is nearly identical to the spot price of the very same NAND FLASH chips used on the inside. In other words, the extra controller IC inside the microSD card is sold to you “for free”. The economics that drive this are fascinating, but in a nutshell, my suspicion is that incorporating the controller into the package and having it test, manage and mark bad blocks more than offsets the cost of testing each memory chip individually. A full bad block scan can take a long time on a large FLASH IC, and chip testers cost millions of dollars. Therefore, the amortized cost per chip for test alone can be comparable to the cost of silicon itself. To ground this in solid numbers, suppose a production-grade memory tester costs one million dollars. If you take one million dollars and divide it by the number of seconds over a five year period (a typical depreciation lifespan for such equipment), the equipment “costs”$0.00634 per second. Thus, a thirty second test costs you $0.00634/second x 30 seconds =$0.19. This is comparable to the raw die cost of the controller IC, according to my models; and by making the controllers very smart (the Samsung controller is a 32-bit ARM7TDMI with 128k of code), you get to omit this expensive test step while delivering extra value to customers — I love the fact that when I put on my linux kernel hacker hat, I can be completely oblivious to the existence of bad blocks and use mature filesystems like ext3 instead of JFFS2, at no extra cost to end customers like you. Isn’t it fun to connect the dots, all the way from silicon die markings to the linux kernel to end users, and all the businesses in between?

In the end, I’d have to say that both SanDisk and Samsung look like they might be superior wholesale vendors to Kingston for memory cards due to their more direct control of their respective supply chains. Unfortunately, you can’t buy Samsung-branded microSD cards on the retail market, as far as I know — Samsung only sells their cards to wholesalers who then rebrand and/or resell the card, and like Kingston these non-OEM brands may blend their vendors so it’s hard to say if you’re getting the best card or simply a usable card.

### 311 Responses to “On MicroSD Problems”

1. Tony says:

Simply.
Amazing.
I’m just a US consumer, and a minor eletronics enthusiast. I love the work you’ve done here, and would love to read similar works on other (re)branded memory chips.
A lot of work (and writing!) goes into benchmarking and the like, but very little actually goes into testing the quality of memory. It’s utterly astonishing that you’ve done this on your own, and the results speak for themselves!
…you could have an entirely new business reviewing various memory chips. “so it’s hard to say if you’re getting the best card or simply a usable card.”

2. Wolfgang Spraul says:

Fantastic post, just want to say thank you, keep it up!

3. Derek says:

I get the feeling the silkscreen batch is some sort of in joke on the ghost shift.

• Tom Keddie says:

Don’t assume english is spoken, it might be more significant in unicode/multibyte.

• Jared Earle says:

Ghost Shift and counterfeit goods will sometimes have an English-language pun, like SHITINERAND in the Monarch Polfy cheap fake Rolexes. However, in this case A00LF is a typical Kingston serial ender.

4. AgentX says:

Hi Bunny,

Really interesting post. I have something to add: about a year ago I bought a Kingston microSD card from an Amazon marketplace seller. It came with an adaptor that strangely was a few mm too wide for an SD slot. This arose my suspicions so I emailed Kingston, providing them with pictures of the card ( http://i50.tinypic.com/aeb7s3.jpg http://i49.tinypic.com/34edtgy.jpg ). It looks very similar to your sample. I received email confirmation that it was a counterfeit card: “Based on the images provided, I am sorry to confirm that the part is not a genuine Kingston product … The numbers on the card are not our serial numbers. Someone had to imitate them.” It seems like we’re receiving different explanations.

• Anonymous Coward says:

Bunnie bought his directly from an authorized distributor.
Hard to tell someone he bought fakes from your own distributor.

• Tarantino says:

Just because it was an “authorized distributor” doesn’t mean that they’re above getting cheaper counterfit cards and calling them Kingston. Especially in todays smaller margin world.

This blog explains why one of my kingston cards will, and one of my kingston cards wont, work in the same machine. I hope the head of the Totem pole @ Kingston does something about it. I had been buying Kingston exclusively, but now I’m rethinking my decision.

• Bloos says:

These are my thoughts too!In India, I have seen this very often. The so called authorized distributors/resellers selling you a fake. And surprisingly, it has happened to me with Kingston DDR2 memory modules.

great post!

5. Michael H says:

Found your post via Boing Boing, and it was a good read on both the problems you’ve experienced, and a good insight into the electronics market in general and the impacts of cost-sensitive production models and specifically the quality/cost curve. Kind of validates my personal choice to only buy memory products from name manufacturers who ‘stand behind their products’ so to speak.

6. Juho Salo says:

This was a fantastic post. Thank you.

7. Lon says:

Thank you for the informative and in-depth article. I love detective stories like this. Great work!

8. Not only fascinating but a real education. Thank you, Bunnie!

9. Jason says:

Thanks for writing this article. I have to second Tony, I’m an electronics enthusiast and really enjoyed this article. Your note about removing the plastic with acid hit a chord with me. In high school I had to do the old find the density of a penny project where you measured how much water it displaced (I think that was the goal…). Instead a friend and I broke the penny in half and weighed it before and after using acid to remove the zinc middle. We then used the atomic weights of the metals involved to calculate things based on the penny’s original weight and the weight of the remaining copper. Convoluted and cumbersome, it was a lot of fun! Now that I think about it, the fun was the best part as I can’t remember if density was what we were really even after…

Thanks for this great article. The time you put into it is much appreciated and your blog is now bookmarked for future visits.

10. RT says:

Great forensics work! I do a lot of things for my work, and one of them is testing chips. I really liked the detail that you went into.

However, I would implore you to shift your manufacturing to a country other than china. A country that straps women to operating tables and forces abortions in the 9th month of pregnancy is not one in which we should be doing business.

• kaye says:

http://www.npr.org/templates/story/story.php?storyId=9766870

I avoid dealing with china whenever possible. They stand against a lot of things the people in the US are passionate about, and we only use their products for the cost difference anyway.

• G says:

And there are absolutely no reasons to avoid US products.

The whole ‘holier-than-thou’ attitude is a bit too much. If you want to punish the manufacturers you better start in your own backyard.

• B. Mused says:

G,

When we start using political prisoners as slave labor I’ll agree with you. When we start executing prisoners to provide organs for re-sale, I’ll agree with you. When we start censoring the Internet, I’ll agree with you. When people like you. who dissent, start being censored and imprisoned, instead of simply being taken to task by others for your folly, then I’ll agree that we are pretty much the same as China. To be sure, we have a political faction in power at the moment that would love to make us the same as China, but they haven’t won yet.

I have two friends who were standing among the other students at Tiananmen Square when the tanks rolled. Although they survived, they were also systematically persecuted for years thereafter. If their parents had not been well connected in a very large and powerful Chinese company, it would have gone much worse for them.

My wife and I have more or less “adopted” a young Taiwanese woman, currently a university student. She is very much a part of out family, and here in the US, we are very much part of her family. I know the fear she has for the safety of her family living in Taipei if the Chinese decide to take Taiwan. Already they are strongly influencing events there. Her brother is in the Taiwanese military, and at particular risk. But most of all, she is native Taiwanese, not Chinese, and so as part of the minority, in a country long overwhelmed by foreigners, her family is at particular risk for persecution.

[For those who are unaware, the native Taiwanese are ethnographically Malay-Polynesian, and have weathered repeated onslaughts, persecution, and near genocide from deposed and displaced Chinese regimes, as well as being subjugated by the Dutch, the Japanese, and whoever else found Taiwan a valuable spot to occupy.]

Is our country perfect? No, of course not. We make mistakes, and being made up of human beings, some of whom chose evil over good, bad things have been done in our name.

But I will put our overall record of standing for what is good and right alongside that of any other country in the world.

It is those people, like you, who wish to draw everything down to the lowest common denominator, and assert that no one has the moral standing to speak out and criticize, it is those people who are the enablers of the world’s great evils.

So smirk sanctimoniously if you wish. Mock those who speak up, deriding them as outdated anachronisms and nationalist bigots. But I will stand with them, and you can stand alongside Mao, Pol Pot, and Stalin, with your assertion that the US is just as bad. I will stand with a frail, sweet, young Taiwanese woman, and you can stand with the people bent on destroying her, her family, and her culture, a culture of whose existence you are probably unaware. Much as they decimated the Mongolian culture of my friends from Tiananmen Square, turning its remnants into a “state cultural treasure” to be used to promote tourism. Perhaps the now suppressed Taiwanese puppetry will one day have such an “honor”.

You may be tempted to drag out the sins of our dusty past, and to be sure, they exist. But I live in the here and now. I know and oppose the present dangers, and can only ask forgiveness for errors of an earlier time.

I don’t hear China acknowledging the sins of her past, nor do I see her government turning away from the sins of her present, including the oppression of Tibet, the threatening of Taiwan, the imprisonment and torture of Falun Gong practitioners, or the arrest of Huang Qi for the simple act of criticizing the government’s response to the earthquake in Sichuan.

Tell me, how many were arrested for criticizing the government here for the handling of the hurricane in New Orleans, even the ones making the absurd and scurrilous charges that the levees were dynamited by secret government operatives?

G, you are a fool. I just wish that you were a harmless fool. Instead, you are a dangerous one.

• kinnie says:

B.Mused,

Using the phrase “sins of the past” implies that you believe all is well and good in American home and foreign policy. Would you care to substantiate that stance with evidence? Perhaps you should be less quick to call others a fool, until you know that you aren’t one yourself.

• dfc says:

kinnie,

You fail reading comprehension, or are just lazy. Quoting from B.Mused post “Is our country perfect? No, of course not. We make mistakes, and being made up of human beings, some of whom chose evil over good, bad things have been done in our name.”

Clearly they recognize that we are not (present tense) perfect.

• Minion says:

• Someone says:

This is quite a bit off topic, and of poor taste. Social issues should be discussed elsewhere, as bunnie has done a lot of research and deserves your attention.

If ethical/social issues is what you live by, then simply state that “I have issues with china’s ethics, and therefore will not buy from them. You should consider those ethics as well, when deciding to purchase from business from that country.”

Otherwise, please be kind and make a blog or website of your own.

• :/ says:

Is that worse than kids starving to death from overpopulation?

• Bob says:

:/ thinks murdering someone else’s baby will keep it from starving?

That is simply a brilliant deduction. Let’s see if we can think of some way to prevent:/ from using any of our food!

We don’t do that in the US, though – it only works in dictatorships.

• JM says:

Hahahahaha! And the US are completely innocent.

On a serious note, great article.

• JT says:

Uh well, yes. Show me anything even *close* to the american government coming in and forcing an abortion on an american woman in an american city.

The “well, america isn’t perfect” excuse isn’t holding water when the crimes against the chinese people are so egregious. It’s just shocking the world will sit idly by shrugging and saying “well…nobody’s perfect”. Maybe this is how the holocaust happened.

• Someone says:

I believe there are some (not all) that are hypocrites here, including one mentioning 10 fold amount of information which could be summed up in a few simple lines, *cou,B.Mused,gh*

I do agree, however, that you should consider the ethics of the country in which a company resides, including researching if that company supports or rejects those ethical issues. Who knows. They may be on the frontier of pushing the a country out of the bad ethics they are surrounded by.

But each person has a set of their own ethics/morals, and thus should be researched on their own. Here is not the place to discuss those.

• JM says:

And bombing innocent women and children in a country accused of harboring non existing WMD is full of human right, grow up.

11. ken says:

Wow, what a great piece of investigative work! I really like the discussion about the “Ghost Shift” – so that is how all the fakes are made.

12. Joe Zydeco says:

Can’t agree more. Insanely detailed but a real look into an industry that I almost never get to see as a programmer. Thanks for all the detective work!

13. John Lei says:

Thank you great detective work, I have a defective 16 gig SDHC with the same funny D.

14. […] Studios’ blog has been on my regular reading list ever since the first XBox was hacked. The latest entry has just surpassed the level of ingenuity and entertainment value of his original paper on the XBox […]

15. jason says:

Bunnie, you are a smart guy. You need to hook up with one of the bigger MicroSD users in the manufacturing food chain and leverage some of the buying power they have and get access to the QC they get in their supply line. No doubt from there your business will grow.

16. […] http://www.bunniestudios.com/blog/?p=918 posted by admin at 10:23 am   […]

17. wohali says:

Hey bunnie. Interestingly, larger Kingston-branded microSD cards I’ve purchased in Canada and the US (8GB/16GB size) seem to have the lot number silkscreened on the front with the exact same ink as the label itself. Like Sample #5, everything seems screened in one pass, and no, mine don’t have the broken “D” in MicroSD. The back does have the same lot number sticker as your Sample #3, so I suspect they’re genuine (and I’ve not had a lick of difficulty with them!)

Your theory about Sandisk/Toshiba selling A-/B+ parts to Kingston makes sense. Assuming my larger cards have the same stacked CSP construction (unlike your irregular unit), Sandisk/Toshiba might complete all the packaging and send unlabelled cards to Kingston’s fab, who then do as they please – including labelling them with whatever markings they want.

In your scenario, presumably the ghost shifter didn’t bother to setup the part of printer that actually read the codes off of each device before using them to print onto the plastic package, and what you’re seeing are the default values in the machine. I wonder if the microSD fixtures that might be used in such a silkscreener line are readily available, and their code or behaviour might be perused…

18. S.L. Herman says:

This is a bombshell. Awaiting to see if any mainstream tech reporter picks up on what you’ve revealed!

19. […] of America’s retail markets over the murky backwaters of Chinese industry. Probably. [Bunnie Studios via […]

20. Jon says:

Great work! I will with interest follow the ramifications from this detective work.

21. Great post. I’ve been interested in this topic ever since I was sold knock-off Sony Memory Sticks a few years ago. The varying quality you get on these flash cards really runs the gamut.

22. Aris says:

I really enjoyed reading that – thank you!

23. […] transparency of America's retail markets over the murky backwaters of Chinese industry. Probably. [Bunnie Studios via […]

24. […] Does the name Bunnie Huang ring a bell? It might if you were around the Xbox “scene” in 2002, when people from all over the world were on Xbox hacking forums trying to figure out what made Microsoft’s first console tick, so to speak. It was truly exciting—we’re talking Paris in the 50s exciting. Mr. Huang has since gotten a job at Chumby, putting his technical know-how to good use and helping to create the adorable little clock-widget-thing. While in China, Mr. Huang discovered a problem with a number of Kingston-branded microSD cards that were to be used in Chumby production. The problem was that they were counterfeit, or “irregular” in the measured parlance of the blog post. […]

• One other thing to consider with a “thirty second” test is if someone needs to test 1,000,000 MicroSD cards (let’s say Nokia orders a giant batch), that’s:

1000000 * 30 = 30,000,000 seconds, or 8333 hours, or about 347 days

And that’s assuming you’re running the factory 24/7/365. So the amount of overhead this could add seems prohibitively large, unless testing could be done in parallel.

25. Very Interesting. I had some trouble with a 4Gb flash just before christmas.

The external packaging look more like you sample #2. Solid D and laser etch batched code. So do the entries in /sys/bus/mmc/devices.

I put down the issues I was having to to MMC bus contention as I could only trigger them by working the wireless – which looked on the device I was using the card with tho be connected via sdio – and the card hard at the same time. Interestingly though changing to a Sandisk card a lot of this went away.

Now there may be bugs in the linux S32C420 sdio implementation – but it interesting what you say about the kingston controller in hindsight.

Unfortunately I don’t really have the time or knowledge /documentaion of the sdio details to do any debugging – not when I found a Kingston cards works. Fortuantely this is for a one-ff project so I don’t have thousand of cards to deal with.

Have you consider that these ‘irregular’ cards migth not be Kingston and they bought them to get them off the market, this doesn’t play well with the rest of the economics you’ve found though.

26. […] the transparency of our retail markets over the murky backwaters of Chinese industry. Probably. [Bunnie Studios via boingboing] […]

27. racy_rick says:

I’m sure this happens quite a lot. With obfuscation of manufacturers ordering something from very far away and many hands touching it. Add to that low wages for their workers and stuff like this is bound to happen.

Thanks for the great extensive writeup. You should go into forensics!

28. Nathan says:

Whoa, I just checked, and the 512 MB one from my phone is a Kingston with a broken D and A00LF. And that came with the phone!

Further checking and I notice that my (fake) Kingston 2GB cards from dealextreme actually have a solid D and multiple ink colors. Interesting.

29. Martin says:

A fine piece of detective engineering, and a joy to read as well – great job!

30. Matt says:

I get excited when ever there is a new post, you should be writing for Wired! So entertaining and unlike so much of the webs rehashed tech specs. I too thank you for your awesome articles! Seriously your giving William Gibson a run for his money with the ghost shift stuff. LOL love it!

31. joe peacock says:

Respect.

32. Alex says:

Got to your story from Slashdot.org. Thank you for your time and funds used in producing this write-up. It was done very well, and raises some very important questions.

33. Bunnie,

Much like wohall, I have a Kingston branded 4GB microSD card whose lot number (N0160-002.A00LF) appears to have been silkscreened onto the card casing using the same ink as the rest of the label. I also have an unbroken D on my card. Interestingly, the microSD -> SD card adapter that came with it does have a broken D on it, though.

Like your original irregular card, Sample #1, my card has odd values:
name: SD4GB
date: 00/2000
manfid: 0x000041
oemid: 0x3432
serial: 0x00000c61
scr: 0235000000000000
fwrev: 0x0
hwrev: 0x2

• markus says:

I wonder if your 4GB card, which is apparently from a ‘ghost shift’ or just from a counterfait manufacturer, has the capacity that is written on it. Ever tried to put up to 4Gigs on it?

• Will T says:

C’mon buddy, here’s hoping there’s some sort of email notification for replies…

How do you get this info about the card exactly? I’ve got a Linux box at home but not the technical know how to work it out. My Googling hasn’t produced good results either.

• Will T says:

Never mind, scrolling down success!

34. For what it’s worth, the card above was purchased from a reputable US retail outlet here in the SF Bay Area.

35. I have found similar things with standard flash drives too. One thing is for certain, a silkscreened logo is a poor indicator of the authenticity of the part.

36. murray says:

Great! ANOTHER amazing blog I have to follow!

37. Bob says:

As a CS that just finished a course in computer architecture and digital systems this was very interesting.

38. Turthalion says:

We order Kingston microSD cards from a North American vendor for a WinCE product (the SDC/2GB card).

I have a couple of cards that don’t work in my microSD reader/writer. They match Bunnie’s descriptions and silk-screening–broken D,, lot code of N0181-006.A00LF.

Here’s the really crazy thing: the cards don’t work in the microSD slot on my reader, BUT they work if I plug them into a microSD adapter, and into the SD slot (which is on the same reader). Really strange, as the adapter is purely a connectivity thing, no electronics involved.

Also, the working microSDs show up as “Removable Disk (E)” in My Computer.

The ones with weird markings that only show up via the SD adapter, when I plug those in to the reader, the “Removable Disk (E)” suddenly starts displaying “Kingston (E)”.

I’m going to take photos of these cards and phone my vendor and link him to this blog post. Between this product and another, we bought about 3K SD cards last year, and we’ll be using another 8K this year, so I’m hoping we’ll get some action.

• Turthalion says:

Further to my earlier post, the bootloader on our product spits out some card details, so here are the results.

The two A00LF cards report Manf ID of 0x41, part # of SD2GB. These are the cards that don’t work properly. The manufacturing date and serial # are sensible values–October 2009 and ). Product revision is 2.0. OEM ID is ASCII 42 (0x3432).

The regular working cards have Manf ID of 2 and part # of SA02G. Product revision is 0.3, and manf. date is March 2009. OEM ID is ASCII “TM” (0x544d).

I don’t spit out CID values, but CID is made up of MID, OID, PNM, PRV, PSN, MDT and CRC, as you can see by checking Bunnie’s results.

I’m in R&D, not at the production facility, so I don’t have a huge number of cards around here, but I’m going to test what I’ve got, and then we’ll be talking to the vendor and/or Kingston.

Will keep people posted if there’s interest.

• Jimbonics says:

Interested.

39. Federico says:

Thanks for writing this awesome article!

Cheers,
Federico

40. […] Here’s a cool post (via BB) which digs nerd-deep into a business problem in the electronics manufacturing biz in China, resulting in some neat conclusions about the structure of the microSD market and paying for quality, even of commodity goods like advanced modern hardware. […]

41. Jay says:

Thanks for the insightful detective work Bunnie.

Knowing how hands-on you are, I’m curious how you ensure that there’s no ‘ghost shift’ of chumby internals being done by your manufacturers.

42. […] CalculatorRelated BlogsRelated Blogs on BernakeBernake did not save the economyRelated Blogs on BlogOn MicroSD Problems « bunnie's blogRelated Blogs on blog archivep2pnet news » Blog Archive » The World According to Googlep2pnet news […]

43. jcavanagh says:

Wonderful post, great work.

I am actually an FSE currently working in Korea. I had a casual conversation with a Taiwanese field service engineer who was working on a system next to mine, about a problem I was having with my client, what he said was simply “welcome to Asia”.

Keep up the good work.

44. […] ######### Verizon to allow unlimited Skype calling over 3G starting next month ######## ######### On MicroSD Problems bunnie's blog ######## ######### 7 Things to Consider for Social Media in the Enterprise ######## ######### The […]

45. Andrew S says:

Thanks for the fascinating read. I still remember back to the days when Kingston was one of the most trusted names for RAM. It’s amazing to see what goes into our dirt-cheap flash prices.

Are any of these cards actually made in Taiwan, or is that another forgery?

46. Deltech says:

This is Fascinating. I loved reading this article. I also had problems with some Kingston cards I purchased, and strangely enough, they looked similar to the “irregular” card you pictured here (The Sectioned “D”, and the last four of the serial ended in “A00LF” also).

I, later, confirmed that they were indeed counterfeit cards.

So I believe they are not telling you the whole story. Strange and fascinating indeed.

I’ll be damned!

I have a faulty Kingston card that perfectly resembles your “sample 1” irregular card. Could you please post more details about the /sys procedure in linux so I can output the same data as you did?

PS: Great article!

• bunnie says:

The procedure iss a little bit tricky and machine-specific. You can’t just have the card mounted as a USB drive via an SD – to – USB converter, since the USB converter hides all of the details of the SD card from the host; you need to have a native SD card interface, like those found on many mobile phones or, as it happens to be, on the chumby One.

Once you have the card mounted via a native SD interface, you should be able to find an entry in the /sys directory tree that looks a bit like /sys/devices/platform/cpuxxx-mmc.n/mmc_host/mmc0 where cpuxxx-mmc.n will be platform-specific. From that directory you will see a subdirectory with a format like mmc0-nnnn, which will have a number of files that contain the CID and CSD data structures. From there you can just do cat * (or grep . * if you want prettier output) and get all the card data.

• Mark Cox says:

The SD card slot on my thinkpad laptop worked too with all the interesting details appearing in the ‘/sys/block/mmcblk0/device/’ directory

• Prashant says:

I tried this with linux mint on a dell latitude E6410. I do not seem to find any of these folders.
‘/sys/block/mmcblk0/device/’ or /sys/devices/platform/cpuxxx-mmc.n/mmc_host/mmc0′

Can anybody help me with this

• G_Ford says:

Has anyone identified low-cost adapters or carrier boards with holders for SD cards, mini-SD cards, and micro-SD cards that when plugged into desktop computers’ mainboards will work to read out the electronic card ID data? Hopefully, a free live CD Linux version like the 100MB Puppy Linux, for example, can be used with the adapters or boards to obtain the data.

Two types of adverse events come to mind that should increase demand somewhat for the above types of devices are:

1) Having a list of your cards’ ID numbers to write on police reports should make it more likely to recover them should they be stolen. In addition, should a nefarious person add a card to your collection of cards, you might have a somewhat better chance of proving that the added card does not belong to you if you have a list of your cards’ IDs.

2) After the all-important random write speed tests have been performed and your faster cards have been identified, knowing their ID numbers makes it possible to recover them should they be accidentally mixed with slower speed cards.

Reference – Random write speed testing: CrystalDiskMark http://chucklohr.com/808/

An adapter or board with only a single SD card socket may be less expensive, but it should be able to read the data of both mini-SD and micro-SD cards that are inserted via adapters into the SD card socket.

It would save a lot of time to have the card slots on the front panels of the desktop PCs, but this may increase the installed cost.

Hopefully, low-cost boards or adapters can be found. The following devices might be good candidates:

40-Pin IDE Flash Drive Carrier Board with a SD card Interface
40-Pin Male IDE To SD Card Adapter

It would probably take too much time to read a lot of card IDs with an old laptop that does not have a card slot, but the following type of adapter might work even if the laptop runs slowly from a SD card such that most users will not be satisfied:

Secure Digital 44-Pin 2.5″ IDE To SD Card Adapter

48. SpeedEvil says:

Interesting! Do you have any sources for datasheets of the wear leveling chips?

• Partarion says:

SD controller datasheets aren’t particularly interesting because the only interfaces they present are standard: SD on one side and NAND flash on the other. Internally, they are microcontrollers (as mentioned in the article), but the datasheets won’t go into much detail about their programs.

Bunnie, do you have any idea who made the controllers in the irregular/fake cards?

• SpeedEvil says:

I was hoping for some details on wear leveling algorithms.

• SpeedEvil says:

Sorry – tired – of course you saw that in the original comment. thanks. Reflashing the flash controller would be fun :)

I suspect that it’ll be one of those several page datasheets that omits any description of the program – as you allude to.

49. […] as simply a vendor that re-marks other people’s chips in its own packaging”… On MicroSD Problems « bunnie’s blog. Yikes! Filed under: EE — by adafruit, posted February 16, 2010 at 9:52 pm Comments […]

50. Allen says:

Just because you purchased from a Kingston authorized distributor doesn’t mean they weren’t trying to make a quick buck. It’s quite possible that the distributor knowingly purchased/manufactured a batch of counterfeits/sub-grade and used their authorized distributor status to pass them off as genuine. This is very likely given the aggressive lead times and pricing.

Incidentally, this is probably how counterfeits get into the major retail chains. A retail store buys from an authorized distributor who ships them fakes, which ends up with the consumer.

51. […] interesting story about computer manufactuing and MicroSD […]

52. AlwaysLearning says:

Great work as always, Bunnie!

I hope that this will flush out more people with issues like yours and Turthalion’s. The semiconductor industry is long-past due having a spotlight shone on the extent of “horse trading” that goes on.

53. Neil says:

I was in China in the early part of Dec’09 and was struck by the large amount of Kingston memory products available from various shop vendors at, once you started bargaining, too good to be true prices. So I demurred on buying because it seemed fishy, even though as far as I could tell the packaging was A-1 legit. A bit later in the trip, on the electronics street in Nanjing I noticed Kingston USB Memory Sticks being packaged into bubble packs in a little shop by a teenaged girl and, presumably, her mother. As I say, you could not tell the difference from proper Kingston packaging that you find in North America, but it did leave me wondering. Very happy that you took the time to investigate and publish your findings.

54. […] just read what I thought to be a most interesting article on the SD card […]

55. […] On MicroSD Problems « bunnie's blogFebruary 17, 2010 – Bunnie is a very smart man. Also good at giving some insight into the hardware/semiconductor/memory industry. […]

56. Gloria says:

Tosh/SNDK flash is self testing and self-binning. It only takes four probes per die (IIRC, it’s been a while) to power it up, clock it and move a small amount of data to verify completion of the self-test routine.

hope this is useful info. :) Gloria

57. Bill says:

kingston?Chinese?Hot Political Figures.

58. Alex says:

Interesting. I sometimes wonder if the vendor can tell if the item is real or fake themselves. With the amount of information available you would of thought in this case they would.

I had a very similar experience with branded polo shirts. The brand could not tell if the shirt was real or good fake.

59. Captain Burn Out says:

I ran a small IT retail business for a number of years. The rebranding that goes on within the industry is truly something to behold. I’ve seen every brand rebadged as just about every other brand. Perhaps the funniest was Corsair chips sold as Kingston. These were all through regular wholesale channels so as bunnie says Kingston don’t make any of their own stuff, or at least not very much. The “A00LF” part number is pretty common, it’s just a numbering system & not anything more than that.

Only thing to add is a customer who bought a microSD card off eBay & came to me for a replacement when it didn’t work. This was the fakest of fakes, instead of the gold contact there was just a bit of plastic crudely glued on … lol. He was greatly relieved when my supplied card registered as soon as he popped it in his phone.

So many stories, maybe I should write a book!

60. CFuser says:

rogue worker ≠ rouge worker

Photographer here. I purchased an 8 Gig Lexar CF card from a major US photography equipment supplier a couple of years ago, encountered repeat read errors. After holding onto the card for almost a year, I contacted Lexar and with some resistance I was able to set up a warranty replacement. The replacement card shared all the same outside markings. It also produced the same identical error. I stopped using it and only use SanDisk without problems. I assumed it was a Camera/Card interoperability problem, but after reading this I have to wonder.

61. fiach antaw says:

Any chance of higher-resolution snapshots of the microSD card die? I had no idea microSDs had microprocessor logic in them, it’d be interesting to see if the code is stored using flash/eeprom (in which case it might be possible to rewrite) or a mask or OTP ROM (in which case it might be possible to read the code off of the die, which would be interesting).

• bunnie says:

Unfortunately, I have no high res images of the die. Flylogic is the place you’d need to go to get something like that, I don’t have that capability in-house yet.

Each controller IC has its own unique quirks, but based on what I saw the ROMs generally looked OTP (I agree reading them out would be interesting). However, the Toshiba controllers used fuses that you could visibly read out with the microscope (the blown fuses had discolored and developed black marks where they were fused), and the Sandisk controller had a region that looks similar to EEPROM, probably for storing the CID/CSD data. Hard to tell, most of it is covered in CMP tiling, though.

Agreed it would be interesting to explore the structure of these ICs in greater detail, if I had the time. I also have a sneaking suspicion that just accessing the raw FLASH data itself could be revealing. There’s no reason why the controllers couldn’t also store some code on the FLASH device itself. Presumably, the bad block maps and other interesting bits and pieces are already stored on the FLASH device.

• afcagroo says:

The blown fuses are probably not extremely interesting. On a controller chip, they are most likely used to disable test modes or to disable the ability to write the controller’s internal flash/OTP memory. If you were to repair them (using FIB) AND you know enough about the controller, you might be able to read its internal program (OTP, ROM, or Flash). This is a much better way to go than trying to optically decode the program from its ROM, as there are multiple hurdles to that route. You are correct that the mapping of bad blocks, wear levelling data, etc. are likely stored on the NAND Flash chips themselves.

• Randy says:

I’m familiar with USB memory sticks and assume SD cards are similar. USB sticks contain a controller that loads its microcode (firmware) from the attached flash memory chip. The microcode is located in the first few blocks on the chip. Most flash chip manufactures certify that the chips they provide do not contain any bad blocks in this area.

The bad block and other details is sometime referred to as metadata and is typically stored on the flash chip (at the logical end of the memory range). Not all controller chip manufactures follow this convention.

62. paul says:

“Kingston wouldn’t take the cards back because we had programmed them”

Could you point me at some documentation on how you program the cards?

• Partarion says:

programming is otherwise known as writing an image to the card using an ordinary card reader :)

• paul says:

I was assuming that the “Samsung controller is a 32-bit ARM7TDMI with 128k of code” meant that it was actually programmable.

• bunnie says:

I was referring to programming the bulk FLASH, not the controller IC. The FLASH part of the card had been programmed with a regular SD card adapter.

• paul says:

Terminology misunderstanging.From the other comments I see that the firmware is ROM. I assumed you would have used “written” instead of “programmed” to refer to actually writing the flash. Thanks for the clarification.

63. Cex says:

Impresive !
This post is fascinating.
Could you please post some more details on decapsulation process and the microscope used? You get amazing pictures of dies.

Thank you for sharing this with us!

Regards.

64. […] checks that were traced to their MicroSD mem­ory cards. Specif­i­cally, they were traced to Kingston MicroSD devices with odd-looking mark­ings. His inves­ti­ga­tion turned up a lot of inter­est­ing and wor­ri­some infor­ma­tion about […]

65. Matt Burkhard says:

Great blog entry!
We need more great info like this on the net.
Thanks for the great, thorough and in-depth write up and experiment!

66. KW says:

The “D” on the SD logo is supposed to be “broken”. Two black streaks give the letter an illusion of reflection that would appear on when looking at hard drive platters. You can check this up by checking official any official images of the cards or retail packages, even on the right hand side of the second image on this page. I suppose the card on the right on the first picture is the “irregular” one, or rather result of sloppy printing.

• Sean says:

I think everyone here knows that. The point is that the official Kingston cards have a solid D, and I don’t believe it’s “sloppy printing” so much as a design decision.

The rationale would be because in design, it’s best to omit tiny features on something that’s already tiny, in order to improve being able to visually register the object. Tiny features can make an already tiny object look confusing, like that feature was a mistake instead of on purpose.

In this regard, it would make sense that the official cards took this into account because Kingston has proper designers. The unofficial cards were made by some guy who tried cloning the original images by using official SD logos off the web, which include the broken D marks.

67. […] Shared On MicroSD Problems. […]

68. Jeff says:

This is the longest article which I’ve read with rapt attention in quite some time. Fascinating.

69. […] interesting story about computer manufactuing and MicroSD […]

70. […] just read a fascinating story about investigating a bad batch of Kingston MicroSD cards. Some of the things I […]

71. Vdrumpro says:

This is why apple keeps sd slots off their handheld devices..(among other reasons, such as thickness) who wants to pay to support and trouble shoot inconsistant storage problems? By NOT offering memory expansion externally, the overall costs stay down and we get blessed with such things as $199 iPod touches and$499 ipads. Step back and put on your 2005 eyeballs.. $499 for THAT? Oh we’ve become so jaded so fast.. Year before last a fake 16gbsdhc class 6 (was really a class 2 rebranded,) ruined my Xmas vids.. Lesson, buy memory from amazon.. Not eBay.. 72. Eoan Kerr says: Love the post, great investigation and writeup. I’ll need to dig out my old microSD cards now, curious now :) 73. Mij says: Fascinating article full of nice detective work. Thank you for sharing your findings. 74. […] READ MORE… Comments (0) Trackbacks (0) Leave a comment Trackback […] 75. fruitfly says: Check this latest EETimes article – Fake parts threaten electronic market: U.S.: http://www.eetimes.com/news/semi/rss/showArticle.jhtml?articleID=222900688 76. XxionxX says: This is great work. I know that I will be looking closer at my media in the future because of this post. 77. Mark says: Really incredible work. Thank you for such an enjoyable read. I just stumbled on it through Google News. Have never read such an in depth article about arcane tech minutiae (and it is relevant to me, the tech-naive electronics consumer). Passed it on to other family members to read. Bravo. 78. […] dollars of scrap, not an amount his San Diego start-up could afford to “sneeze,” he wrote in a blog post detailing his dive into the rabbit hole of Chinese high-tech […] 79. Not Anonymous says: > Unfortunately, you can’t buy Samsung-branded microSD cards on the retail market You can, at least in Germany. http://www.reichelt.de/?ACTION=3;ARTICLE=97197;GROUPID= 80. […] very interesting article on bunniestudios regarding the issues with some allegedly rebranded MicroSD cards. Geeky and […] 81. feroze daud says: This is an amazing article! I learn a lot about SD memory cards, manufacturing and supply chain. Also, I was under the impression that fakes are actually manufactured in separate factories, not as a ghost shift in a name-brand outfit. I had a high opinion of Kingston, now I will doubt all their offerings, including PC RAM and flash drives. 82. Twirrim says: Fascinating. In the past I’ve always trusted Kingston as I perceived them to be a quality brand, guess I was wrong on that front. I have a Sandisk 8GB microSD card in my N810’s external slot, the OEMID comes up as 0x0000, but I guess that’s probably down to the adapter? 83. […] On MicroSD Problems « bunnie’s blog – Fascinating detective story: how bad microSD cards are sold through second and third tier vendors: (via @gnat) […] 84. […] February 18, 2010 Every once in a while, someone writes a really great detective story. […] 85. […] of America’s retail markets over the murky backwaters of Chinese industry. Probably. [Bunnie Studios via […] 86. The photo of a “generic MicroSD card” in the Wikipedia article on MicroSD shows a broken D: http://en.wikipedia.org/wiki/MicroSD 87. […] On MicroSD Problems « bunnie's blog […] 88. […] On MicroSD Problems « bunnie’s blog Speaking of the Über-geeky, what do you get when you cross a Hardware Engineer with a Detective? Bunnie…P.I.?!!? Read how Andrew “Bunnie” Huang used his M.I.T.-honed skills to uncover a mystery — and be thankful that some people know what the hell they’re doing in this world. Somebody get that man a mustache, eh? (Via http://www.bunniestudios.com ) […] 89. Ah, This is perfect! Puts to bed many contradictions I’ve seen 90. You wrote: “I love the fact that when I put on my linux kernel hacker hat, I can be completely oblivious to the existence of bad blocks and use mature filesystems like ext3 instead of JFFS2, at no extra cost to end customers like you.” I disagree: http://cananian.livejournal.com/58238.html How do you know you’re not getting subpar controller code on those sketchy dice? And the algorithms used can never be updated or fixed once the SD is made. I’d prefer to have my flash fully exposed, thanks. • bunnie says: That’s an interesting link, and a good point. I do think that clever people could do a lot to optimize FLASH filesystems if it was exposed to them in some way (I ponder this possibility in the comment thread of a previous post about my SSD, where I did get a bad controller that eventually crashed and I lost all my data – http://www.bunniestudios.com/blog/?p=592 (see comment thread, it’s not in the main body of the post)). However, despite my general agreement with your notion, I would have to counter with a couple of counter-points. 1) I have to ship a product today, and the fact is that JFFS2 and YAFFS pretty much suck, despite the heroic efforts of the developers behind them. This is the whole chicken-and-egg problem, as you point out: nobody sells a reasonably priced “raw” plug-in card device and therefore nobody develops for it and in the end we end up with x86-like abstractions (they do exist in droves, however, in embedded platforms). While I do take on philosophical challenges in the course of business, I don’t take them all on at once, and as practical matter using ext3 + invisible translation layer got me to market faster and better because I was able to spend more development cycles testing other features that users really care about, like battery management, instead of fighting with an unreliable filesystem that eventually leads to bricked units and/or extremely long scan/wait times during boot. 2) Getting a FLASH translation layer generically right is *very* hard. It’s not just about the filesytem layout. It’s also a lot about ECC, wear leveling, and matching disparate logical sizes to physical erase block sizes — and the reality that as densities go up, the physics of these devices is increasingly coming into play. For example, as bit-densities go up we become increasingly dependent upon ECC and the ability to properly detect when a block is about to “go bad”…simply writing to and reading back might not be good enough, there are subtleties like how long you wait between the write and read to verify the integrity of the data. This sort of requirement for intimate knowledge of the process points to an argument for a better integrated and matched controller-to-FLASH system, rather than a decoupled one. In other words, as memory starts looking more “analog” in terms of bit-errors, you want to push the translation layer into the “PHY” and keep the linux involved only with the “MAC”. As a note, in rotating media there is a boatload of magic firmware in there that we don’t see. I’ve seen some of what goes on inside the hard drive firmware, and it is Black Magic. The servo heads actually trace a non-circular pattern as the disk goes around to compensate for minute flaws in the bearings, for example. Could a driver in linux handle that? Nay. Really, hard drive companies have become more firmware companies than anything else. Platters, heads and servos are commodities, but the real magic is in the DSP algorithms used to error correct and control the servos, and to manage bad blocks. Nobody has batted an eye at the boatload of code contained inside rotating media, and ext3 is just fine for it; I think that FLASH is just starting out on this path, and will eventually follow this route. To elaborate, the typical error I have seen in existing FLASH filesytem implementations is they assume certain things to be true across all FLASH devices — like the size of the OOB field, the format of the ECC, or the size of an eraseblock, or the method for marking and replacing bad blocks. Unfortunately, every chip has a different way of doing these things, and even a very skilled coder will have trouble navigating and validating against the myriad of possible FLASH configurations. It’s also a thankless task, to wit — if you make a mistake, a lot of people will be very upset because most people take a reliable filesystem for granted. While you are right that a subpar controller with terrible code on them will do bad things for you, that is part of the reason I’ve opened up the devices, to see what the pedigree of the controller is. I haven’t gotten to the point where I can read out the code in the controller to analyze its algorithms, but I’m looking for matched controller/FLASH manufacturers on the theory that the closer you can get to the quirks of the FLASH device the better chance you’ll have of optimizing correctly for them. Hence, my general inclination to buy a Sandisk or Samsung card with matched controller to FLASH, versus a “mix-and-match” of controller to FLASH. To be clear, this *is* a primitive notion, but so far the only notion one can have without access to NDA material. As the market matures, and our understanding of the market matures, a more sophisticated method for judgment may arise. That being said, I think it is a worthy goal to try and build a really robust, cross-platform FLASH filesystem — it could solve problems and bring about higher, more robust performance. I’m just saying that this is by far too hard a task for me, and in the meantime while smarter people work on it, I’m going to avoid using it, and inspect my memory cards very carefully and try to pick the best ones I can. And, there are still a lot of raw NAND devices out there for developers to play with who want to take on this challenge, mostly in mobile phones, routers, set top boxes, and even the classic chumby. • Donnie says: Out of 100’s of thousands of devices, I have never shipped an embedded (uC)linux device with a writable filesystem. I make the whole FLASH only writable as whole filesystem(s), and then use name=value tags to generate configuration files in tmpfs on the fly. Writing to a filesystem on FLASH opens the device to the possibility of fs corruption. I romfs/cramfs/squashfs everything including config file templates. If I need installable packages, they are written either a fixed offsets if NOR FLASH or to a FATfs on serial FLASH or SD. In normal operation, -linux- never writes to flash, ever. Config changes go through the Bootloader layer, and it just appends a few 10s of bytes to a config log area with whiteout entries. Philosophy, I guess. • Russ says: You should checkout UBI and UBIFS. Really takes a lot of the lessons learned from JFFS2, YAFFS, and newer NAND flashes. There is also some thought about putting btrfs on top of UBI. Really makes me want to get my hands dirty and put together a PCI card with an FPGA and NAND flash and start experimenting with a hardware accelerated UBI implementation. • wookey says: This is a fascinating discussion. I am a fan of access to the real flash and filesystems designed to deal with that (I’ve been involved with YAFFS since its inception). I am distrustful of the idea that the (incredibly cheap) flash controller in my SD card will do a better job than a properly written driver and filesystem. This is particularly true because that controller is written exclusively to do a good a good job of running a FAT filesystem, mostly writing large contiguous flies as fast as possible (i.e the camera/media player use case), because that is still 99% of the market. That means it only track 2 ‘write areas’ and it keeps extra bad blocks for the file allocation table at the front of the device. That’s terrible for ext3, where the journal is at the end of the drive and you need more than 2 ‘write areas’ in use at once. The firmware in the controller is not at all interested in being good with ext2 or ext3 or btrfs, and Arnd Bergman has done some realy good characteristation recently of the problem: http://lwn.net/SubscriberLink/428584/98a701f81ad3263e/ If we could talk to the flash directly then it wouldn’t matter that the manufacturers only cared about FAT and cameras. Sadly that’s not the way the market is going. SD flash is now _much_ cheaper than solder-down ‘real’ flash, and the solder down-stuff is getting built-in controllers too: ‘eMMC’, so it seems that raw-flash filesystems are not the way of the future. Which is OK, but they do need to work properly, and right now many of them work very badly indeed in the sort of use cases we are interested in. Reliability is terrible, speed is poor. You do make a good point bunnie, that the disk drive people have done a great job of putting a lot of caching and firmware between you and the drive platters which does all sorts of magic and we’re not finding that a problem, but we have a long way to go until the SD people get their act together and do such a good job that their algorithms are filesystem-neutral, and we still get reasonable lifetime out of the flash. It would be great if we could talk to the controller manufacturers/firmware writer people and get cards made that worked better with real filesystems. After all – it’s just an ARM7TDMI so writing some code for it is no big deal. YAFFS would just about fit in that 128K :-) (that probably isn’t the best plan, but it would be really interesting to try). I have no idea how we get to talk to the right people about this – the use of real filesystems on SD and eMMC is only going to grow and we need to get beyond the ‘FAT is the world’ view many still seem to have. The controller does have some advantages over a linux driver in that it has better bandwidth to the raw flash and more info on manufacturing/performance details. But it also has the major problem that the power can be yanked at any moment, and what you should do about that depends a lot on the use-case, which is why UBI has a lot of tuneability in terms of how long you are prepared to wait before hits the real flash. Collaboration and good work in this area is no doubt hampered by a load of crappy patents that will have been filed on ‘anything sensible you can do to improve flash error rates and performance’. That probably makes the firmware authors nervous about opening up their code. 91. Found Objects says: Found on eBay… have to wonder if this isn’t somebodies brother of the person running the ghost shifts. All Kingston, all high enough sizes to raise eyebrows. http://shop.ebay.com/dealz_central/m.html?_nkw=&_armrs=1&_from=&_ipg=25 92. Kevin says: I’m suprised by your suprise that kingston doesn’t make its own chips/SD… you never see kingston listed on EEtimes as a big consumer of semiconductor maanufacturing/tooling • Kevin says: oh, and great post! very informative. • bunnie says: This is an element of the post which was poorly communicated; allow me to clarify. It is well-known to many (including me) that Kingston is not a semiconductor manufacturer, in that they own no fabs, hence my overly-brief conclusion seems a platitude. However, simply not owning a fab doesn’t mean that you don’t “make chips”, or you don’t collaborate closely with a fab. nVidia owns no fab but is casually referred to as “making chips” with their logo on it; few bother with the tongue-twister of “the chip fabricated by TSMC containing nVidia IP” in favor of “the chip made by nVidia”. I was expecting for a big player like Kingston to at least have a chip on the inside with their marking on it, although of course made in someone else’s fab. They invested in 3S, for example, a company that designs NAND flash controllers (it may well be that the controller chip marked 3023C is in fact a result of this partnership, but it is not exclusive to Kingston branded chips as you can see from the decap results — samples 1 and 4 have the same IC). So when I said that Kingston is simply re-marking other people’s chips, I’m not being surprised by the fact that they don’t make their own NAND FLASH or own any fabs. Rather, I was expecting a Kingston-branded or at least unique to Kingston controller IC on the inside — some sort of special Kingston “value add” that would give me a real reason to buy their technological solution over someone else’s. At the very least, for the volumes shipped it would seem to make economic sense to own your controller ASIC IP (as Sandisk has done) instead of buying it from someone else. Still, “value add” need not lie specifically in silicon IP. It could also have come from innovative packaging technology or a multi-source supply chain position that can deliver cost advantages via leveraged buying power. However, in a (perhaps too small) sample of 4 chips, Kingston exclusively used Toshiba/Sandisk ICs, yet Sandisk is a direct competitor. I would have expected some more diversity, say, incorporating Samsung, Micron, Hynix, Spansion…someone other than Toshiba/SanDisk…in their cards. Being able to balance the supply chain and be less dependent upon a large competitor for a supply of chips would in itself be a significant value add to customers, since that gives Kingston leverage to negotiate a better price that few others can achieve. Presumably, Kingston isn’t buying the Toshiba/Sandisk memory chips from Sandisk, but rather from Toshiba — still, as far as I can tell it’s the same chip and same fab. In my mind the potential for conflict of interest is staggering and I’m very curious as to how this all exactly came about. How could Kingston get a better deal out of this than Sandisk? There is nothing in this arrangement that would lead to a self-evident conclusion that, for example, a Kingston card with a Toshiba/Sandisk memory chip and a Toshiba controller should be superior or cheaper to a Sandisk card with a Toshiba/Sandisk memory chip and a Sandisk controller. The situation of buying a significant amount of a competitor’s technology from a competitor’s fab yet still selling at a competitive price is counter-intuitive (at least to me), and perhaps my greatest folly was simply expecting something different. Unfortunately, the press picked up on these short, controversial but unsubstantiated soundbites and many knowledgeable people people have slammed me, drawing the conclusion that I am utterly naive about the semiconductor market and thereby discrediting the work. Which is fine, I deserve it for not choosing my words more carefully; I was personally more excited about the technical details and failed to attend to what I regarded as incidental stream of consciousness observations and not really the meat of the post. As you can see from the length of this response, I had substituted about a page of thoughts for two (poorly phrased) sentences; it is the constant struggle of a writer to espouse brevity without forsaking clarity or completeness, a struggle on which I had failed. • Stu says: Bunnie, This is a great article and I’m sure eye-opening for a lot of people. However, I thought it was common knowledge that the whole reason Kingston came into existence was to act as a demand buffer (as you pointed out) by taking excess DRAM production from the likes of Nanya, Powerchip and focus on turning these into modules and selling them (while the DRAM makers focused on the memory manufacture itself). As I say, though, great article. Thanks! • Bloom Berg says: Beeing the largest DRAM module supplier in the world and controlling 38% of the market (according to iSuppli) I would be inclined to think more of Kingston as an important market leader and not a ‘demand buffer’ relying on excess production. It would be interesting to note that the second largest supplier is A-Data with 8% of the market. Playing on the flash memory market is logical for Kingston due to its well established position with suppliers, customers, and advertisers. But at least my Kingston DRAM modules have Kingston branded chips on them. • Hamza Ahmed says: They also give people another option when buying chips (ie. thumb drives, microcards ie. – SD, memory duo). Its ok, everybody makes. What matter is that you actually admitted. Trust me, a lot don’t 93. mark says: Fair dues to you for doing this article. Was a very enjoying read 94. Mr.pro says: I’m Korean and I’m a high school student so, my sentence can make nonsence. It is so impressive I’m not good at english . I’m have difficulty in interpreting. and I can misunderstand something, but through your investingation , I can explore I have’t and I’m surpriszed that responces are longer than your article I’m begining my blog recently, so no responces in my blog… 95. CapnKernel says: Regarding URLs to validate chips, it is entirely possible for the chip manufacturers to digitally sign the chip serial numbers. After all, the chip has storage! If they did so, you could go to their website and look up that chip. If it hasn’t been registered before, you’re good. If there are thousands of registrations for that signed serial number, you know you have a cloned chip… • JT says: I like it. The question is–do the manufacturers really want to draw attention to what is apparently a HUGE problem? I know I won’t be buying Kingston any time soon. 96. […] (permalink) link: On MicroSD Problems bunnie's blog Not pointing any fingers, this post has quite a but of interesting stuff on the state of grey […] 97. Peter Freiman says: Hi Bunnie What a marvelous detectives work you have done! :0) Very interesting to read – an thank you for spending your time and money on this subject. Best Regards Peter 98. […] On MicroSD Problems « bunnie's blog (tags: cool interesting electronics hardware flash engineering memory microsd kingston) […] 99. […] Huang – whose name you might remember from inside the chumby One – was prompted to investigate an apparent bad batch of Kingston microSD cards when the touchscreen widget device (which stores […] 100. M says: Bunnie, I’m curious about your economic analysis. The addition of a controller chip might eliminate the cost of testing the memory chips… but doesn’t the manufacturer then have to bear the costs of testing the controller chips? 101. Embedded says: Bunnie I believe Kingston’s value proposition is to provide a working ram to the consumer through their web site which is guaranteed to work forever. Kingston has for example the Pentax K20D DSLR Camera listed on their web site. Through the strength of their reputation I bought two 4GB SDHC flash (marked China and with suspect lot numbers) from Fry’s in Houston Texas. They did not work. Kingston replaced them with 4GB Japan (Toshiba) SDHC Flash cards. Now I 102. Embedded says: I meant to say that to an OEM such as yourself the Kingston value proposition is not useful. Quite frankly you are better off with Samsung, Spansion, or a group such as PNY / Patriot that assemble only good die to your specification. While I am in Oilfield and will buy Toshiba product with Industrial Temp spec and low defect (and can pay for it) I would buy to data sheet that can be used for incoming inspection for your part. (Perhaps Look for Industrial Temp to start with get the correct data sheet then go commercial for yourself. This works for Freescale!) 103. […] Red Ferret Journal | Bunnie Studios | Email […] 104. […] Red Ferret Journal | Bunnie Studios | Email this | Comments Share News Engadget for iPhone / iPod […] 105. […] Red Ferret Journal | Bunnie Studios | Email […] 106. […] Red Ferret Journal | Bunnie Studios | Email […] 107. […] (bunniestudios via redferret) […] 108. […] Red Ferret Journal | Bunnie Studios | Email […] 109. […] Red Ferret Journal | Bunnie Studios | Email […] 110. […] Red Ferret Journal | Bunnie Studios | Email this | Comments Posted in Tech News Tags: microSD, production, […] 111. Awesome post! I’ve known for a while Kingston just sell-on under their brand. I bought a Kingston 2GB SD card a while back, and I could see something underneath the sticker on the front. I yanked it off and it was just a Toshiba SD card with a Kingston sticker. It must’ve been A-/B+ as it was cheaper than the same card by Toshiba. 112. […] Red Ferret Journal | Bunnie Studios | Email this | Comments ArrayArrayArrayArrayArrayArray Link To This Post1. […] 113. James Wong says: Now lets hope discovery or bbc picks that story. Should be interesting to see footage from an hidden camera during those ghost shifts. 114. […] Red Ferret Journal | Bunnie Studios | Email this | Comments This entry was posted on onsdag, februar 24th, […] 115. Nelanka says: Excellent writeup! Its great to see stuff like this explained so clearly. I do worry about buying SD cards and the like from some of these Amazon storefront vendors. Good to know what to look out for. 116. […] Engadget, sourced from Bunny, just wanted to past along a really fascinating article around SD card production in Asia. If […] 117. […] Red Ferret Journal | Bunnie Studios | Email this | Comments Go to […] 118. No suprise to me – we’ve seen this same issue with Kingston Compact Flash cards a few years ago – they were even selling cards that weren’t fully compliant with the standard – they wouldn’t work correctly in all modes. We found that their cards that use the Silicon Motion controller were garbage. We now only sell and recommend Sandisk CF cards. 119. […] Red Ferret Journal | Bunnie Studios | Email this | Comments Share and […] 120. Cris says: Great post! 121. Blair says: Amazing post, Bunny. You just made me a fan of both your blog and your company. I might have to buy a chumby as a thank you for the compelling content of your blog. 122. […] EST. Please see our terms for use of feeds.Permalink The Red Ferret Journal | Bunnie Studios | Email this | Comments Tags: andrew huang, AndrewHuang, bunnie huang, […] 123. […] EST. Please see our terms for use of feeds.Permalink The Red Ferret Journal | Bunnie Studios | Email this | Comments none Leave a commentCommentsLeave a Reply Kliknij […] 124. Saturn says: The A00LF lot code seems to show up on a variety of Kingston products. I have a ~3 year old 2 GB Kingston ‘DataTraveler’ USB stick with the lot code 05360-023.A00LF. Listed as Chinese manufacture, rather than Taiwan. • JT says: Yep, me too. I have a 1GB and a 4GB and both have the A00LF and both are from China. Soooo…what’s this mean? They’re fake? 125. […] This is by far the most interesting blog post I’ve read this year. It also gives me a great urge to buy something – anything – made by Chumby. If the guy who runs the company is that on the ball, they must be making good stuff, I reckon… Share and Enjoy: […] 126. […] The sleazy world of MicroSD. […] 127. Tim Brazer says: That was a great post. There’s a lot of stuff in there I never would have noticed. I actually have one of those Kingston SD cards with that A00LF silk screen code on it. Now I know it was made part of a ghost shift. I bought it on eBay about 2 years ago for$.99.

128. Ben Levy says:

This is a real eye-opener. I never would have imagined things would be so complex for a little card.

129. Dave says:

This article was really intriguing. The idea that “ghost-shifts” were happening at such a well-known suppliers plant (and that they were so blaisé about it). The mind boggles.

130. […] Armed with this evidence, we confronted Kingston — both the distributor in China as well as the US sales rep. First, we wanted to know if these were real cards, and second, if they were real cards, why were the serialization codes irregular? After some time, the Kingston guys came back to us and swore these cards were authentic, not fakes, but at least they reversed their position on not offering an exchange on the cards — they took back the programmed cards and exchanged them for new ones, no further questions asked. – bunniestudios.com […]

131. Attrezzop says:

Wow!

Epic post. Blew my mind, it’s the first time I’ve heard about “ghost shifts”.

I’m kinda interested in what the chinese gray market looks like. That in itself is interesting. I imagine it’s a bit like the weekend computer shows here in the US.

132. Jack Hsu says:

Thanks Bunny! Great info!

133. LeeNukes says:

Echoing previous comments, absolutely fantastic. It was very interesting reading and thanks for taking the extra effort of going that one step further.

134. […] On MicroSD Problems Open Standards, Open Source FSF: Why I Will Not Sign the Public Domain Manifesto They Don’t Make Computer Manuals Like They Used To A Floating Office in Every Dock […]

135. if the controller is arm7tdmi, it is likely to be programmed through the sd interface itself. it would be SO cool to reprogram it to do something else…think about it…a tiny tiny ARM with GOBS of nand to play with and 128K of flash and at least 128K of ram (for wear leveling they must keep at least one eraseblock in ram)

any ideas how to program this arm?

136. […] On MicroSD Problems « bunnie's blog […]

137. Great post indeed and very eye-opener… thank you!

138. […] Red Ferret Journal  |  Bunnie Studios  | Email […]

139. pmshah says:

I recollect seeing Kingston ads declaring that the company was doing so well that they gave bonus of some US$30,000 to every employee. I wonder where they stand now. Here in India they warrant their products for 5 years generally on no questions asked replacement across the counter basis. I also recollect an interview of Kawasaki China chief by, I believe BBC, where in he made a categorical statement that they are not inclined to introduce new models in China since a fake (motorcycle !) would generally be in the market in less than 6 months. I also recollect reading about seizure of a large quantity of fake Microsoft software CDs in US docks. The holograms were developed by I believe University of Shanghai. When the state itself sponsors – not only condones – wholesale piracy what can you expect. One of the main reasons why China has not been able to make inroads in the automotive components business involving large scale manufacturing. Practically all the auto manufacturers want their sourcing to be in India. Perhaps because there is too much at stake. To top it all the US consumers do not have the kind of nationalistic feelings like the Japanese. I have experienced this first hand. A Japanese end user preferred to pay a$ more for a Fuji transparency film rather than buy a Kodak one. They want their jobs to be retained in US but do not want to pay the higher cost it entails.

They can’t have both and have to learn to live with it.

140. […] nizka serijska številka teh kartic. Če vas zanima bolj natančna preiskava pa le odklikajte na njegov blog. Več na […]

141. Zrce Urlaub says:

thanks very useful

142. Alap says:

Is there a simple Utility which checks whether the Flash Storage is Genuine OR Not.
I am aware it will not say: “Sorry Buddy, your Kingston 2Gb Card is Fake!!”.
But it can at-least say: Data Read / Write Speed = XXX MB/sec. Reference Card Read / Write Speed = XXX MB/sec. Your card seems to be of inferior grade.
Eg: Usual Read Speed is 16-20 MB/s & Write Speed is 3-6 MB/s. If you get 7-10MB/sec read speed & 1-3 MB/sec Write Speed, your Card is Definitely of inferior quality &/or Fake….

143. Randy says:

Very interesting post.

The comment on the ghost shifts producing counterfeit flash cards confirmed my suspicions about the source of false capacity memory products (USB sticks, SD Cards, etc) that are sold on the internet and appear to originate from production facilities in China. The numbers of cards that are available for sale and sold indicate the ability to produce these fake products in volume and thereby access to production facilities.

Kingston is a popular brand that appears on internet (Ebay and other markets) with false sizes. These fakes are also referred to as “upgraded”, “oversized” or “capacity adjusted”. A internet search on fake flash drives will indicate the magnitude of the problem.

I’ve done some preliminary investigation into how the memory cards programming were altered to report a false size and my conclusion were that the techniques used to modify the firmware on the cards varies by controller type. Also, in the case of one controller all that is required is the modification of the metadata on the drive.

What I find concerning is that there are indications that some of the flash memory devices do not report memory errors to the operating system (in the case of windows a media error 30).. I’ve repaired some cards with the manufactures flash programming tool (restored firmware and rebuilt metadata) and multi-bit errors were not detected. I’ve confirmed this by using the program H2testw to test the flash media.

145. […] On MicroSD Problems – “So I kicked into forensic mode. The first thing that raised my suspicions is the external markings on the irregular Kingston cards. ..” […]

146. […] “Bunnie” Huang, co-creator of the Chumby smart alarm clock, posted an interesting exploration of the underside of MicroSD memory card production in China.  Huang came across a batch of […]

147. […] post on Kingston and their MicroSD cards. Unfortunately, looks like all the 2GB MicroSD cards I have are […]

148. […] + info – bunnie´s blog Categorías:Nokia Etiquetas: microSD, Varios Comentarios (0) Trackbacks (0) Dejar un comentario Trackback […]

150. pete says:

Hi Bun et all;

I use a lot of memory cards. I work in the imaging world and always look for faster write speeds. I ran into problems recently with some fast Sandisk cards. One of the high end cameras wouldn’t write to the cards. Some testing found the camera to be okay and I called Sandisk. In speaking to a sales engineer he told me that about 1/3 of the sandisk cards on the market were forgerys. He said that the copys had seriously hurt their business. He gave me a list of trusted vendors. A couple of weeks ago I called Signetics to ask about an older analog product. The engineer I spoke with directed me to an up-dated chip and I asked who stocked the chip. He told me to buy it on their web page. I could buy one unit if that’s what I waned. He said that the problems with copies in the market had produced a solution.. buy directly from us. This looks like a solution. It’s likely more profitable for the Mfgr. pete

151. pete says:

I guess I’m showing my age, It wasn’t Signetics.. I miss Signetics. It was Intersil. pete

152. Yes, the counterfeits must be eradicated – and we have the product to do it.

Now the Manufacturers have to step up to the plate…

We are working on it :)

http://swiss-authentication.us/default.aspx

• pete says:

That system is unique to my knowledge. It looks great. pete

153. Amy says:

I finally decided to write a comment on your blog. I just wanted to say good job. I really enjoy reading your posts.

154. […] unusual markings on the chips, he decided to investigate for himself, comparing the ID data and dissolving the cards’ casings with nitric acid to take a look inside. He found that each of his Kingston-branded samples actually had a Toshiba/SanDisk memory chip […]

155. Mohan says:

Damn!
I thought Kingston was trustworthy …
Great work ..

156. CPRM DRM says:

There’s a mess of DRM in those embedded flash controllers (“Secure” Digital contains crypto DRM stuff designed for Hollywood: CPRM, Content Protection for Removable Media). So: reading and reverse engineering the embedded firmware will be useful for figuring out how to cripple the DRM so Hollywood can never use SD cards’ hidden modes against us.

An example of an SD “card controller” company is Hyperstone.com, a German company that has made several generations of SD card controller chips. Some of these chips boot from the flash media itself, so by probing the flash and reading it directly, you can read their firmware. See their support documents.

MMC cards are compatible with SD but leave out the DRM crap. But somehow what’s cheap in the market is SD cards with CPRM built-in to screw us.

157. […] Bunnie’s Blog: On MicroSD Problems – “Sample 6: This is a SanDisk card bought on the open market from a sketchy shop run by a sassy chain-smoking girl who wouldn’t stop texting on her mobile.“ […]

158. I agree and I will come back and follow more of your posts.

159. […] had read about fake flash devices before on Bunnie’s Blog, so I started doing some searching online to see if anyone had perhaps a simple way of finding out […]

160. it is interesting and informative article. This has been very helpful understanding a lot

of things. I’m sure a lot of other people will agree with me.

161. […] packaging, but… counterfeits nonetheless. Sandisk, Kingston, Lexar, you name it they fake it: On MicroSD Problems bunnie's blog remember: if the deal seems too good to be true, it probably […]

162. […] o&#118er t&#104e &#109urky &#98ackwater&#115 of C&#104ine&#115e indu&#115try. Pro&#98a&#98ly. [Bunn&#105&#101 Stud&#105&#111s v&#105a […]

163. Aaron says:

Very interesting read. I can see you have a passion for the tech as well. I have been wondering about counterfeit and fake memory cards for a while… I new they would surface. I once purchase a 512k USB flash drive (back when they were more valuable than today) and it would not save more than about 128k of data. It was a no name brand with a company logo printed on it (like having “chumby” put on it as a promotional). It could of been just a bad unit, but mine was not the only one. Always wondered if it was bad or ? BTW, love my chumby…

164. […] Pun aside – In October 2009, I’ve got a new scope, a Tektronix DPO4034. It’s a scope in the $8k range, so while not exactly low-cost, you don’t need a mortgage on your home to buy it. The raw specs are nice, but not awesome – 2.5GS/s, 350 MHz, 4 Channels, not much more than the$2k-range TDS2024 that I was using before. But while the TDS2024 has a sample depth of 2.5k, the DPO4034 has an awesome sample depth of 10M! That means that you can actually sample once, and then spend time on analyzing this measurement. Or store the measurement, and do proper postprocessing. Suddenly, the scope stops being a multimeter and starts being a universal all-purpose tool. The DPO-series all have a number of software options, delivered as small dongles that you can stick into some slots in the scope, that enable certain software features – features like UART, SPI, I2C, CAN and even USB protocol decoding. The scope runs ppc44x-Linux, but that’s really just for the frontend. The backend is a number of Tektronix ASICs, that manage the high-speed data acquisition and visualization. […]

165. […] tear down of SD cards and fakes. It was really interesting and has some insights into the industry: On MicroSD Problems bunnie's blog Those are nice boards on your blog Jason. That clear PCB material is pretty cool. Mark […]

166. […] came across this brilliant bit of persistent forensic work over on Bunnie Studios […]

167. Diane says:

Great post! This will help us to know the common problem of microsd.I’m interested with this article.Thanks for sharing your great article to us more power to your site! G-d bless ;-)

168. Diane says:

Great post! This will help us to know the common problem of microsd.I’m interested with this article. And Find solution to it thanks for sharing your great article to us more power to your site! G-d bless ;-)

169. john mark says:

It is sad to know that a good manufacturer produced a good quality product but due to counterfeiting it destroy the image of the original maker. Why not pursue legal action to stop the counterfeiting world wide. I as user, I was a victim of fake micro-sd. Thank you of sharing your article it give us idea to determine the original one. More power.

170. tz says:

I recently had to check the erase structure (ACMD 13) and found a class 6 card (at least on the front) reported itself as class 2.

I’m going to be working on an SD verifier.

171. […] has become much more sophisticated in the past two decades:  Hardware engineer Bunnie Huang performed similar a similar autopsy when investigating a run of bad memory cards from […]

172. Eddie says:

Ours is a Compact Flash card issue. Up until 2009, we had no problems with CF cards. Now, without any notice, some manufacturers have switched to a different (maybe cheaper) controller. We use the card in IDE ATA disk mode using the FAT32 filesystem and we don’t write, only read. The older cards work fine, but the newer cards appear to scramble data when we read a multi sector cluster of directory data. We first noticed the problem with Kingston cards. Kingston has exchanged our quantity of cards and told us that it may be a compatibility problem with the new CF controller chip. Other cards are starting to show problems, Ridata and Transcend tells us they switched to a new controller. Anyhow, we have put up a list, which can be seen here: http://www.cdadapter.com/faq.htm#FLASH

173. […] link to a blog that makes for interesting reading. especially if you prefer kingston memory cards: On MicroSD Problems bunnie's blog __________________ "anyway, it's not rocket […]

174. […] micky found. it makes for interesting reading. especially if you prefer kingston memory cards: On MicroSD Problems bunnie's blog __________________ "anyway, it's not rocket surgery". Last edited by Evil […]

175. Ruben says:

Great and very interesting post. We need more post of a high quality like this one has in the internet.

176. This website just made my month!

I’ve been searching around for talk about this. I’m feeling good now that I ran across this webpage.

Super site!

177. Albert says:

Nice entry: neutral investigation, technically stunning (i am not an expert). The most i liked your saying at the end “I love the fact that when I put on my linux kernel hacker hat, I can be completely oblivious to the existence of bad blocks and use mature filesystems like ext3 instead of JFFS2, at no extra cost to end customers like you. Isn’t it fun to connect the dots, all the way from silicon die markings to the linux kernel to end users, and all the businesses in between?”

Only point for me to critique: no or weak conclusions from all that hard, good work. You are the expert –> what is your conclusion? Your recommendation if any? What is the learning in one or two crisp statements?

178. History Today is following the trend this month, with a reassessment of Alexander’s personality by one of our leading historians of ancient Greece, Paul Cartledgc. But we are at the same time, I hope, helping lead the trend, with a well-argued hypothesis on one of ancient history’s thornier mysteries: the whereabouts of Alexander’s tomb. Our contributor Andrew Chugg even ventures a step further, concluding with a plausible suggestion it can be no more than that, unless DNA analysis is done as to the whereabouts of the young conqueror’s body itself. Were he proved right, it would offer a high-profile example of the common habit of early Christians to adopt and adapt pagan imagery for their own purposes. And it would be fascinating to see whether the discovery of this famous body transformed the modern obsessive cult of celebrity into a medieval-style cult of christian louboutin discount .

179. chanel says:

This is a good step toward where we should be heading.Open source on all levels.

180. herveleger says:

Thank you so much! This is SO helpful.

181. toward says:

This is a handy posting, im happy I found this

182. I agree with everything that was posted in this article, I’m a loyal follower so please keep updating so frequently.

183. Comfortabl y, the post is really the sweetest on this notable topic. I match in with your conclusions and can eagerly look forward to your upcoming updates. Simply just saying thanks will not simply be enough, for the phenomenal lucidity in your writing. I will directly grab your rss feed to stay abreast of any updates. De lightful work and much success in your business endeavors!

184. […] Here is a great article on quality of micro SD cards (in this case Kingston). On MicroSD Problems bunnie's blog […]

185. […] Originally Posted by TurdFerguson Yea it does look a little generic, I just jumped on it when I saw the link. The price has since gone up a bit but when I get it I will post what it looks like, if it matches pic, etc Well check out this article in case: On MicroSD Problems bunnie's blog […]

186. ben says:

Check out some counterfeit chips involving ATmega328s:
http://www.sparkfun.com/commerce//news.php?id=350

187. guess says:

On MicroSD Problems bunnie s blog very nice, This article write well. I has been collection it. 0817/08/08

188. […] of America’s retail markets over the murky backwaters of Chinese industry. Probably. [Bunnie Studios via […]

189. markus says:

Would be interesting to make some investigations about the 16GB and even 32GB Kingston micro SDHC cards that are sold on wholesale plattforms like dhgate, aliexpress and finally ebay. They seem to have similar details as yours above but they often carry faked chips that say it’s 16GB, but when you look a bit further its only 1 or 2GB.
I don’t really think that most of those 16GB micro SDHC cards are real Kingston cards… wouldn’t be good for their business…

190. just_me says:

Great! A must-read for all the folks who insist to buy cheap SD cards in chinese online stores and think they make a bargain…just to complain when their data are lost after a short time of use. ;-)

191. zhanghe says:

i thank i can

192. Tereza says:

I checked now card in my mobile and it is fake one!?? How they can do that? I am from Croatia and probaly it is easy toput fake on our market :-(

193. How can any company have such bad customer service. I just say on the phone for 30 min

194. munky says:

*SIGH*

Kingston IS a re-branding company. they take a whole load of different cards from other manufacturers.

Not to mention they might use a newer printing process.

195. […] Be careful. There are *so* many fake microSD cards out there. Even lots of "legit" retailers end up with them. On MicroSD Problems bunnie's blog […]

196. Kingston seems not so durable, why not use Sandisk Ultra and stuff.

197. Jack says:

I’m buying Kingston products and I had to read this post. It’s kind long, but very informative.

198. Joe says:

Moral of the story: Don’t Fuck with the Bunnie!

199. Rajiv says:

I live in Dubai and this market is flooded with fakes for memory cards particularly microSD- from the brand to the capacity. We also have a problem of upgraded cards vs genuine – I can easily relate to what you have mentioned. You notice huge price swings due to this.
One question – how can you differentiate between the ‘upgraded card’ – in this card the memory for a 2GB card will show as 1.8-1.82Gb which is standard but you cannot open any files from the phone. Will appreciate your reply on this.
Thanks
Rajiv

201. kevin says:

how can you differentiate between the ‘upgraded card’ – in this card the memory for a 2GB card will show as 1.8-1.82Gb which is standard but you cannot open any files from the phone?

of the darkness

202. Thank you for your post and pictures about On MicroSD Problems…I will tell my friends to come here to learn…good

203. Phil says:

Sad that I’m reading this post now, but it makes sense, and people should take heed. I sit here scowling at my “16gb” MircoSD card that corrupts everything past 2gb….the ebay seller is more than willing to send me a new one when I pay to send this P.O.S. back, and ironically it costs me almost 2/3 of my purchase price to send it back….lesson learned :)

204. I had experienced buying fake MicroSD for very cheap price. Believe me it’s not worth it. Go for original one.

205. Singaporean says:

Absolutely fasinating work here. A*Star should hire you, perhaps one day you may like to consider migrating to Singapore and contributing your amazing talent to this small but dynamic country who is aspiring to transform itself into a world-wonder? We need people like you to contribute to Singapore’s knowlege export! Simply amazing, you.

206. demo says:

Fantastic findings. Now I most go through all my memory cards and be pissed off for days about getting taken for a ride with fakes. I will want to tell my friends but they will probably care less… “well, it works… that’s all I care about” they’ll tell me. Oh well.. thanks again!

• PepsiHog says:

First to make me laugh!! LOL

And how TRUE! Paranoid much now?

207. PepsiHog says:

Enjoyed the blog! Great info! :)

As for all the other stuff……keep going at it. It makes for some interesting reading! We will NEVER all agree. (except never agreeing, ofcourse.)

208. jcn50 says:

Keep up the good job, very very informative post & perfect investigation!! Cheers.

209. mike bloom says:

Thank you all for the great readings, my comming into the world of micro sd cards cost me $500.00 sent to a scammer in china. I hope that this cost can lead to a great outcome. I am still searching for a supplier here in the us to ship to miami. I need to buy an average of 200 to 500 cards bymonthly. I would be very thankfull if anyone can direct me to a vendor. my mail is allnews at excite dot com Thank you 210. TomA says: Our company had a bad experience with “fake” chips. Our board used a radio-frequency chip from a reputable manufacturer, for which there was no second source. All of a sudden, our boards started failing thermal test. Whole batches of production (our volumes are not large) would fail thermal test. We narrowed the failures down to this particular chip, but we couldn’t figure out why they were failing. After all, there was only one source, and Reputable Manufacturer must have tested these chips, right? And the chips were old, nearly obsolete, low volume, and didn’t cost that much to start with ($2), so no one would have reverse-engineered them to become a second source — it wouldn’t pay. What was going on?

So we called a failure analysis company, and gave them a few good (old) boards and a few failing (new) boards. They noticed (we hadn’t) that the chips on the failing boards were considerably different from the ones on the good boards. The marking was sloppier, the encapsulation mold was different, etc. Some of the bad chips even had visible bond wires (!) peeking out of the epoxy. They decapped the chips and found that the bad ones generally had corroded metallization on the silicon — more evidence of a bad encapsulation job. But the silicon dies had the Reputable Manfacturer’s logo on them. And the failure analysis guys told us what probably happened, although of course they couldn’t prove it.

Reputable Manufacturer had a testing and packaging line in the Far East (they all do). Someone on that line was pocketing dies (naked chips) that marginally failed test — or possibly even chips that passed, while reporting a higher-than-real test failure rate to his bosses. Then he got together with his buddy who owned another packaging line, and they did a midnight run with the marginal/stolen dies. And another buddy snuck them into the reputable supply chain. And we, on the far side of the world from all this, got stuck.

211. Hernán Peralta says:

Awesome information. I’m suffering these fake microSD cards, but our problem is the high power comsumption that they have in IDLE state, from 0.200mA (not bad, almost 30% of the pieces) to 192mA!!!!! (really hot jejeje). Almost 10% drain 30mA.
Cheers from Argentina and sorry for my poor english.

212. I was fairly pleased to get this website.I wanted to thank you for this very good read!! I clearly enjoying each and every small bit of it and I have you bookmarked to verify out new stuff you post.

213. Wholesaling has become a dangerous business with sites like dhgate being chalked full of scheisters taking peoples’ moneyand never communicating again. When web surfers start researching wholesale, they’re usually the first resource discovered. Great read though.

214. Jim says:

I purchased 3 “SanDisk” 32GB MicroSD cards through ebay sellers located in Hong Kong. They all appear to be counterfeit. They sort-of work (not reliably) in an adapter in my computer, but they will not work at all in my EVO phone. ebay is not a lot of help in dealing with this.

215. infinityxxi says:

I sometimes wonder if the vendor can tell if the item is real or fake themselves.

216. Gary says:

I too had purchased a Kingston 32gb micro sd card from Ebay that doesn’t work. Just got it yesterday. Almost identical markings as your irregular one. Business was in Florida I believe. I will check /sys to get the details. Time to start emailing PayPal, Ebay, and the retailer to try and get my money back….

217. mariusz says:

I`t not nice to read about Polish people stereotipe. I know that in whole world aren nic word abou us but in Poland left nicest people

218. […] Looking into this as well. Found this article that discusses some of the fakes on the SD card market. Interesting: On MicroSD Problems bunnie's blog […]

219. you are really a good webmaster, you have done a well job on this topic!

220. Ferres Art says:

Do you have twitter account friend? So i can track your website

221. Ayzek says:

I too had purchased a Kingston 32gb micro sd card from Ebay that doesn’t work. Just got it yesterday. Almost identical markings as your irregular one. Business was in Florida I believe. I will check /sys to get the details. Time to start emailing PayPal, Ebay, and the retailer to try and get my money back….

222. Gary P says:

I too had bought a 32gb micro sd from a seller on Ebay from New York. Same lot number and such, although mine works in my HTC Evo. It’s seen as 32GB, so I’m gonna put it through it’s paces and load it up to ensure it can handle the data, not to mention I’ll check to see if it can keep up with music/video.

223. jimmy choo says:

Jimmy Choo’s beginnings can be traced back to his workshop in London Borough of Hackney, North London, which he opened in 1986 by renting an old hospital building.

224. coach says:

Coach handbags and purses are made from the highest quality leather and are manufactured by highly skilled workers that take pride in the products they make.

225. Nice sharing, love the information. All the best from Hamburg

226. Amazing to see that so many persons and companies experienced fake chips. Can’t believe the costs involved.

227. Trein Parijs says:

Would be nice to get an answer on one of the earlier questions e.g. “One question – how can you differentiate between the ‘upgraded card’ – in this card the memory for a 2GB card will show as 1.8-1.82Gb which is standard but you cannot open any files from the phone.” Anyone has a solution for this problem.

228. osman eralp says:

I just ran across this post, and it explains a lot. I have two Chumby One units. The Kingston 2GB SD cards in both units failed within just a few weeks. Fortunately, when the first one failed, I could dump an image from the second one and put it back in the first one, with a new SD card, of course. I swore never to buy another Kingston SD card. It’s great to see an explanation for what I experienced. –Osman

229. antoine says:

Hey I work in a La source store in canada wich is an electronics retailer with over 700 stores in Canada, so I believe our kingston cards are legit, and if you look at the kingston 8gp micro SD class 4 package with a micro SD adapter, the adapters SD also shows these markings, and the SDs printed on the back too

230. Nice post. Thanks a lot.

231. […] that Microsoft said they would not support this upgrade, and that its own testing, verified by others, suggests that this upgrade is fraught with reliability and performance issues. But AT&T has […]

232. Pelf says:

Amazing.
Thanks for the insights.

233. Big Scam says:

Memory investigator!

Awesome post, keep up the good work.

234. It’s nice to be informed. I’m using Micro SD as well, here in our country, Phil., there are many “knock-offs” being sold. It’s good to know that mine is the regular one, I’m not aware about this or what the difference between the two ’til I read your post here. Thanks :)

235. Hamza Ahmed says:

Thankyou so much Bunnie! Your work is greatly appreciated and quite inspiring. Your post has greatly expanded my knowledge about tech and helps make a more informed decision and recommendation to all. Keep up the excellent work Bunnie.

236. Mikey boy says:

A company’s margins, stocking, speed, costs, are not really much dependent on the objective, inherent costs of the components which make up their product. It has a lot to do with company culture, logistics, scale, etc.

So when you write: “Oddly enough, of all the vendors, Kingston quoted with the best lead times and pricing — better than SanDisk or Samsung, despite the competition making all their own silicon and thereby having a lower inherent cost structure.” you assume, wrongly, that Sandisk and Samsung has a lower inherent cost structure, as you did not really consider other important matters which determine true costs save that of perceived component cost. In fact, it should not be surprising that Kingston’s cost for flash may be lower than that of the wafer manufacturers themselves – but Kingston supplies liquidity to these fabs by fast payment, massive purchases, etc.

237. PCB Assembly says:

As an electronics manufacturer, I can say that counterfeits are a real problem — the only solution is to buy from the big distributors and bear the additional cost that involves.

238. Andy says:

Totally agree with PCB, fake cards are now getting to be a real issue. The problem is that the general public is unaware of fakes and of course ebay is full of very genuine looking fake cards. We need to make the public more aware somehow of how much a false economy it is to buy fake SD cards.

239. Hey I work in a La source store in canada wich is an electronics retailer with over 700 stores in Canada, so I believe our kingston cards are legit, and if you look at the kingston 8gp micro SD class 4 package with a micro SD adapter, the adapters SD also shows these markings, and the SDs printed on the back too

240. jack says:

Nice stuff I want to buy a micro sd card this will help me out. thanx

241. […] get much help from UK based consumer groups but it might be worth contacting Which? Also, read here. It's very appropriate and very informative regarding fake cards. __________________ Every forum […]

242. Jason Volpe says:

WOW. What an awesome post that got overshadowed by politics. Many people read tech blogs to get away from the politics of today, I know I do. Sure, I got sucked into reading the comments, but once I read about the 9 month old abortions in China, I can’t help but feel it’s a slap in the face to Bunnie who clearly made a significant investment in the research above.

Bunnie, great job on this!

243. Gebze says:

Hi friend did you’r had other articles ? I relish article develop so me … ..

244. Prashant says:

Wow,
That was a very nice post. Currently I am facing disk access problems with my Sandisk disks. Hope they are original and not refurbished.

245. Dan Ruiz says:

This was a simply amazing article! Had my attention from start to finish! You’re also VERY considerate of consumers which definitely hits me deep! Due to this, I’m gonna go with a SanDisk over a Kingston regardless of Class speeds. Your blog is DEFINITELY one that I’ll be checking frequently. Looking forward to reading more of your articles!

246. Calvin says:

Great article. I found this well written and, most of all, well executed.

Makes me question a Kingston card that I purchased from eBay for a steal. It worked for about 3 weeks and gave up the ghost. Not thinking that it could have been a fake or bad stock, I contacted Kingston and they were quick to send me a replacement that worked.

Thanks for taking the time to do this public service. I’ve learned a lot from this article including why Kingston cards seem to get a bad reputation on Newegg for not meeting speed class claims. I will never look at their company the same. Having said that, I just purchase a WinTec 16GB Class 10 card with Made in Korea stamped on it for only \$30. I was a bit aprehensive about getting such a good deal, but this article makes me believe that this is a Samsung NAND. The card tested at 11MB/s write speed so I’m happy so far.

247. Thanks for this post, I became aware of fake ones, but I admit I have some difficulties identifying the authentic from fake ones. Thanks a lot for this very informative post.

248. […] the image above to this one (from Bunnie Huang‘s excellent article On MicroSD Problems) shows that the crack ran through the area where silicon lies. This photo, where I finished […]

249. rand says:

Great article but you’re painfully bias against Kingston, it drags the tone of the post down and you pull the wrong conclusions from it all.

250. superultra says:

A “thesis-grade” piece of investigative -journalism. Outstanding.

I commend you, sir.

251. thanks for the research, i’m bookmarking this site and signing up for the feed.

252. doodles says:

very interesting read, thanks for sharing!

253. Army Strong says:

I just purchased one of these irregular cards off eBay from seller idobiz09 (John Paladino, idobiz@yahoo.com). It looks identical in every way to your sample #1 picture. Needless to say, I could not get it to work properly no matter what I tried. The seller wouldn’t refund or exchange, but I got a refund by filing a claim through eBay. I wanted to see if anyone else was having trouble with the same card so I typed in the lot number and found your article. Thanks for doing the research and posting your findings.