[Ken Shirriff] is no stranger to Hackaday. His latest blog post is just the kind of thing we expect from him: a tear down of the venerable 8008 CPU. We suspect [Ken’s] earlier post on early CPUs pointed out the lack of a good 8008 die photo. Of course, he wasn’t satisfied to just snap the picture. He also does an analysis of the different constructs on the die.
Ever wonder why the 8008 ALU is laid out in a triangle shape? In all fairness, you probably haven’t, but you might after you look at the photomicrograph of the die. [Ken] explains why.
He also explains a bit about how PMOS works and the history of the design, including why it was in the odd 18-pin package. At the end, he talks about how he decapsulated the part and got the pictures, in case you ever want to try that yourself.
As a personal aside, I used to do this at Motorola and I think [Ken] was wise to stick to the ceramic packages since you can mechanically decapsulate them. With an epoxy part, you can use a Dremel or similar tool to mill out some epoxy (just don’t go too deep), put the chip on a hot plate (a copper bar helps carry the heat up to the package), and then fill the milled cavity with fuming nitric acid. But you shouldn’t be doing that without a lot of protective equipment including vent hoods, safety showers, and experience storing, handling, and disposing of nasty acids. I have a feeling [Ken] could pull it off, but it isn’t something you just want to try on a whim.
[Ken] has done this kind of thing before. If you are wondering what kind of computer you could build with such a tiny device, we just saw one the other day. Of course you already saw [Ken’s] talk about his process at this year’s SuperCon, right?
So disappointing! I clicked on the link after seeing “BOOB EXPOSED”. LOL
Would that make this a “chip slip”?
I recently wanted to decap a chip. I googled and found there are companies that will do it for you. I paid $136 for two chips. Well worth it to avoid buying, using, and disposing of fuming nitric acid.
Will they remount them, and is that extra? I have a few chips I’d like to play with bare but I have no way of reattaching bond wires.
Someone I know used NMP to decap a chip, a solvent that’s in a lot of graffiti removers but it’s now understood to be toxic to the reproductive system.
I didn’t ask about remounting.
Did they delayer in addition to decap? Did they take die photos? Don’t keep us in suspense, who were they?
It would be interesting to compare with the 4004 “slice” chip that was claimed to be the doubled for the 8008.
From TFA, the 4004 and 8008 were completely different architectures that had nothing to do with one another.
When I grow up, I want to be like Ken Shirriff!
Motorola decapping the competition. ;-)
I worked for Datapoint in 1995-7 and repaired many 2200v2 and 5500 processor boards and a handful of the old serial 2200’s. The 2200v2 is like an 8080A running at 8 MHz. The 5500 was a Z80 running at 20 MHz. The instruction set makes a lot of sense in octal notation and the change from 8008 to 8080A was to prevent them from directly running Datapoint’s code.
The 2200 had 121 SSI and MSI TTL chips. The 5500 had 131 chips and some were 74S series.
That should be 1975-77
It’s OK, we figured it out. :-)
Very good stories. I imagine the harder part is going from photos to schematic, unless there’s an automated way.
I worked for a company that would reverse engineer chips. We would send them to a company in China that would photograph the layers and turn a team of 500 engineers loose on it until they produced a schematic. They used software to partially automate the process, but it was still a very time consuming process.
As best I can remember from a conversation many years ago with Marcian “Ted” Hoff at Intel, the chip was laid out by hand with Rubylith tape on Mylar sheets. Quite a process.