Counterfeit Metal Oxide Semiconductor

August 22, 2017 7:37 AM

Ken Shirriff examined the internals of Intel's first chip (a 64-bit RAM), including taking his own die photos. Fresh off this, he came across die photos of a different 64-bit chip and knew one of them was not what it seemed.
posted by a snickering nuthatch (29 comments total)

That's both an engaging detective story and an intro to computer chip design and operation.

Don't skip the footnotes!
posted by notyou at 8:00 AM on August 22, 2017 [5 favorites]

It's also a bit enraging if you actually use IC's to, you know, build stuff.
posted by Bringer Tom at 8:01 AM on August 22, 2017 [2 favorites]

One of my hobbies is vintage computers. Counterfeit chips are definitely a big problem when trying to source replacement parts to repair the things.
posted by fimbulvetr at 8:05 AM on August 22, 2017 [4 favorites]

It's also a bit enraging if you actually use IC's to, you know, build stuff.


Yes, all you want to build is a simple DAC and you end up with a missile control computer. Aggravating every time!
posted by Laotic at 8:29 AM on August 22, 2017 [2 favorites]

What's the motivation in the counterfeit market? As Shirriff says, this is hardly a high-demand or high-price part. Why go through the trouble of passing off a DTMF generator as a RAM chip? It's not like this is a fake USB drive that still sort of works, the moment someone tries to use it, it's clear it's a fake.

My best guess is someone had a surplus of the DTMF chips, realized they were the same package as the RAM chips, and figured it was a way to get rid of them. They probably made hundreds of dollars.
posted by Nelson at 8:39 AM on August 22, 2017 [2 favorites]

The IC stuff was way over my head, but I was interested to learn how touch tone (DTMF) dialing phones originally worked.
posted by straight at 8:48 AM on August 22, 2017 [1 favorite]

What a great write up, crazy that someone would go to the trouble - this is clearly why most of my half baked diy projects never work...
posted by zeoslap at 8:49 AM on August 22, 2017 [3 favorites]

I think that's right - DTMF chips must have had huge production runs compared to many others in the 1970s and 80s, so there will have been large stocks of unused parts on the surplus market when the handset makers moved on to a new part. If you have hundreds of thousands of effectively obsolete chips, you'll sell them on for a few cents a pop to anyone who'll take lots of them.

Somebody with access to a chip packaging facility after hours could make useful money by buying the DTMF chips in bulk and relabelling them to a wide range of chips with the same package - and it's a very common package. Doubtless they pushed out many different kinds of 'TTL' devices, not really caring how popular each would be - someone would buy them, and the input cost would always be much lower than the sell-on price, even if that would be twenty or thirty cents. You're not paying for the relabelling (well, you're probably paying someone off for access, but the cost of the machinery and consumables will be absorbed by the legitimate owners of the plant, whether they're in on the scam or not), you're paying cents for the surplus stock and you're seeing margins of many hundreds of percent when you flog them into the chain where they'll soon be lost in the noise.

It's a nice deal, if you don't get caught, and unless you're very unlucky you won't be.
posted by Devonian at 8:58 AM on August 22, 2017 [2 favorites]

I was interested to learn how touch tone (DTMF) dialing phones originally worked

If you should happen to be stuck in the (shockingly distant) past: replace the 3.579545MHz crystal with a 6.5MHz crystal to make free calls from pay phones.
posted by uncleozzy at 9:06 AM on August 22, 2017 [6 favorites]

It isn't just ICs. Other electronic parts are also being counterfeited. Counterfeit capacitors are a big problem.
posted by fimbulvetr at 9:07 AM on August 22, 2017 [2 favorites]

this is hardly a high-demand or high-price part

Not now it isn't, but forty years ago it was. ICs were (at least) a couple bucks apiece.
posted by neckro23 at 9:11 AM on August 22, 2017

The Occam in me says that manufacturing just f'd up and printed the wrong identification on a batch of chips.
posted by achrise at 9:24 AM on August 22, 2017 [2 favorites]

DTMF already.
posted by Ogre Lawless at 9:37 AM on August 22, 2017 [27 favorites]

Ogre Lawless wins, this thread is OVER.
posted by Annika Cicada at 9:40 AM on August 22, 2017

Not now it isn't, but forty years ago it was.

I assume the DTMF dialer chip was also expensive 40 years ago considering the first one was introduced in 1975.

To answer "why," I think Nelson has it: someone with a load of these next-to-worthless touch tone generators and some way to cheaply relabel DIP-16 chips decided to make a little money by counterfeiting.

This is interesting stuff and reminds me of bunnie studios which has more of this sort of thing (previously), in particular this post and this post.
posted by exogenous at 9:41 AM on August 22, 2017 [5 favorites]

I was trying to remember the other hobbyist inventor, thanks for pulling bunnie studios up, exogenous.

(And Ken has a lot of interest in bitcoins, after looking over his archives.. )
posted by k5.user at 9:45 AM on August 22, 2017

Hey buddy, I got this great source of chips real real cheap, fell of the back of a truck, just let me know what you want, cash only, one time deal, grab a bunch, one free with every big bag'o'chips.
posted by sammyo at 9:45 AM on August 22, 2017

I got this great source of chips real real cheap, fell of the back of a truck

How much are they? Cheap as chips, of course.
posted by uncleozzy at 9:49 AM on August 22, 2017 [1 favorite]

Less than 15 years ago a couple of small towns I worked in had trunks that could be blue boxed and payphones that could be red boxed. Also had a lot of folks still using party lines(!)
posted by wierdo at 9:50 AM on August 22, 2017 [1 favorite]

I got this great source of chips real real cheap, fell of the back of a truck

Hey! My friend Bob Sacamano...
posted by lagomorphius at 10:17 AM on August 22, 2017 [1 favorite]

I had an interesting mental block reading this... let me explain:

I always figured a counterfeit chip was perfectly good and functional chip from a second source relabeled as the primary better sources product. Like knock off purses, clothing, etc. It never occurred to me it wouldn't actually work.

Once upon a time I had a CPU chip that was a bit wonky, and then the cap fell off and it turned out the 486DX50 was actually a 486DX33.... so I guess that was a "counterfeit" chip?

It worked once I adjusted the clock speed... for years.
posted by MikeWarot at 10:32 AM on August 22, 2017

Counterfeit radio frequency power transistors are also a big problem, because the prices of some of those are high enough to justify people making the weird packages some of them come in. Others are now no longer made, but come in standard packages, and RF power transistors fail often enough in service to keep demand high. Although people also fake things like 2N3055s, which aren't expensive - but they go to the trouble of mounting a working (but much smaller and lower-power) transistor chip inside the packaging so the fake part passes low-power go/no go testing, only blowing up when you actually use it.

I don't know whether it's true for fake ICs, but knock-off transistors frequently use cheap ink for labelling, so a standard test is to see if that comes away with isopropyl alcohol and a little elbow grease. If you look on Youtube for 'fake transistors', you'll see a lot of horror stories.
posted by Devonian at 11:11 AM on August 22, 2017 [3 favorites]

I always figured a counterfeit chip was perfectly good and functional chip from a second source relabeled as the primary better sources product. Like knock off purses, clothing, etc. It never occurred to me it wouldn't actually work.

They totally can be. Sometimes they're even the same dang thing by the same dang factory. I can think of three broad categories of counterfeit:

feature fraud: the specsheet suggests one set of features, and the package contains something completely different. I don't imagine this is terribly common, but is fairly straightforward. Enabled by market reputations -- you can only defraud a customer once, but if there's enough marks out there, could be viable. Or in some cases, I guess it could just not work at all.
brand fraud: not made or designed by the brand it purports to be, but is fairly similar in design, and likely of inferior quality. Essentially free-riding on another brand's reputation.
ghost shifts: was designed by the brand it purports, and made by the factory the designer contracted with, but was made without reporting to the designer, or delivering it to them. To avoid detection, they often blank out or reuse serial numbers embedded in the chips. Ghost shifts free ride on both the designer and their brand, and often on their factory owners, depending on the level of complicity.
posted by pwnguin at 12:15 PM on August 22, 2017 [3 favorites]

This is great fun. (Including the metal loops digression.) I look forward to reading more of this blog.

I'm betting on achrise's mislabeling explanation. Selling enough intentionally mislabeled chips that can't possibly work to pay for the cost of creating a legitimate-seeming business and establishing reseller contacts before the first end user complains that all 10'000 ICs failed QC seems like a gamble. But, I wasn't buying ICs at the time, so I could well be missing something.

Also, why didn't people label their silicon in the past? There's plenty of room on that die for a model number.
posted by eotvos at 1:57 PM on August 22, 2017 [1 favorite]

Also, why didn't people label their silicon in the past?

Seems pretty likely in the early days that nobody expected these things to be reverse engineered in the field. They were entombed in ceramic and epoxy and that was that.

Once it was found out that decapping was going on, interesting messages appeared...

Lots more fun chip art at the Silicon Zoo...
posted by JoeZydeco at 2:22 PM on August 22, 2017 [5 favorites]

> The Occam in me says that manufacturing just f'd up and printed the wrong identification on a batch of chips.

Making knockoff DMTF chips doesn't seem like the sort of thing that Texas Instruments ca. mid-1970s would be doing.

One of the commenters in the link guesses that a third party was counterfeiting DMTF chips and deliberately disguising them for the benefit of a knowing customer. If there were active patents being evaded, especially if the chips were intended for export, that could make sense.
posted by ardgedee at 2:44 PM on August 22, 2017 [7 favorites]

Jpfed, I first read this thread hours ago and I just got the CMOS joke in the title. Slow clap.

(sorry if this has been obvious to everyone else for the past 12 hours...)
posted by Joey Buttafoucault at 7:24 PM on August 22, 2017 [3 favorites]

wierdo: "Less than 15 years ago a couple of small towns I worked in had trunks that could be blue boxed and payphones that could be red boxed. Also had a lot of folks still using party lines(!)"

I think party lines are still relatively common in areas with large Amish populations - there's no point in paying for a dedicated line when you use it so infrequently.

When I moved to Lancaster County, PA in the mid-90s, the woman at the phone company asked me if I wanted my own line or a party line. "...You mean, like they had on Green Acres or whatever?"
posted by Chrysostom at 2:39 PM on August 24, 2017

Long story short, I hand scanned our local office and found a few loops.

THEN I ended up in Junior Achievement. ( "Hey, Hey, Throw It Away, 'cause It Was Made In The USA!" )

which was in my hometowns CENTRAL OFFICE after then went to the Unix boxes. ( Since they had all this open room where the old switch used to be... )
posted by mikelieman at 2:50 PM on August 24, 2017