Before going deeper into the analysis of today’s chips, we will take a quick journey to where it all began: the Intel 4004, world’s first widely-used microprocessor. The 4004 and most other antiquated chips differ from modern chips in two main characteristics: They only use a single type of transistor (PMOS or NMOS) and each logic gate is custom-designed to best utilize the available area — an inevitable optimization for chips built from transistors about 150x larger than those used in their modern descendants.
The pictures below show four custom-designed variations of the same logic function, 2-NAND:
Each of the gates is composed of two transistors and one resistor. If either of the transistors is open (that is: having Vcc applied to its gate), the output is strongly connected to Vcc. If neither of the transistors is open, the gate is weakly connected to GND through the resistor, but still strong enough to pull the output to GND. The next image shows the only metal layer of the 4004, just above the 2-NANDs:
PMOS is very area-efficient, but more power hungry and slower than alternatives such as CMOS, which combines PMOS and NMOS transistors as illustrated in this post. It’s beautiful to see how none of the inefficiencies we see in modern chips are found on the 4004 and how the available space is completely filled with logic. The entire 4004 has only some 2,300 transistors and makes for a perfect exercise in learning neat chip layout and logic gate design (click for a high-res version):
[edit - Jan 9, 2009: Adding mosaic of entire substrate]
(Clicking on the picture above will result in a 45 MB download!)
As a challenge for next time, identify the extra 3 layers that the Intel museum claims. Last episode’s challenge was correctly solved first by Jeri Ellsworth. Respect for her almost perfect circuit diagram as well as her remarkable on-your-kitchen-table semiconductors fab.
Credit for the chips go to Tim McNerney. Tim is an expert on the 4004 who has built an interactive exhibit of the chip for the Intel museum. For more information please visit the Intel 4004 35th anniversary project web site.
-Karsten Nohl
I’m not sure which two layers we ‘have’, so here’s my guess on all 5: (rot13)
Zrgny
Pbagnpg (Z1 gb CP/Npgvir)
Cbyl
Npgvir
Ubyrf va gur Biretynff (sbe cnqf)
Very nice pictures. How big is the actual die?
Ah good question. I’ll measure it when I have time. We need to update this post with top metal image as well. The die image posted is missing the metal layer so that we could show you the gates.
[...] silicon wizards at Flylogic have certainly posted an interesting chip this time around. The Intel 4004 was the first widely used microprocessor. The logic gates are much larger than [...]
I did not find 3 extra layers, although it does have 5 mask layers with buried contact between the diffusion and poly layer.
The stackup is like this
http://www.flickr.com/photos/jeriellsworth/3183965520/
1. diffusion/field oxide cut
2. buried contact cut
3. poly etch
4. contact cut
5. metal etch
The full process would probably look like this
1. Grow field oxide
2. Cut active areas
4. Grow gate oxide
5. Buried contact etch
6. Poly deposition
7. Poly etch
8. Pre-deposition
9. Diffusion drive in
10. Contact etch
11. Metal deposition
12. Metal Etch
13. Passifying layer
14. Bonding pad cuts
Related to the old chip technology:
http://oms.wmhost.com/misc/
Some of the images a very large.
Very nice images. Kudos for taking time to post this!
This kind of photos shows just how much effort went into creating those first chips… We are too spoiled today with RTL synthesis tools and standard cell libraries
Hi 6581,
Funny you should show us those nice pictures! Some readers sent a C64 asking for that part 2 others to be images and stripped down but we are so busy it’s hard to find the time.
Anytime Andreas!
I really hope those C64s you get are certified to be completely dead. Especially the SID. They are slowly becoming “rare” these days.
Karsten,
That’s an amazing image! We’ve got some great technology for showing off huge images like that on the web:
http://livelabs.com/seadragon-ajax/gallery
If you’re interested, I’d be happy to help you make use of it on your site.
Cheers,
–Ian
I agree with 6581. Due to the ravages of time, and fire no-one has the schematics for the SID anymore and the current FPGA based reproductions are just educated guesses… or am I thinking of the VIC-II?
Either way both chips are very interesting.
If Mr.Admin sees this…
How difficult it would be for you people to decap a working SID (I know I know, it sounds really bad), so that it still works (heat most likely becomes an issue), and then by the use of microprobing, measure several properties of the filter while in use?
Another piece of information that could be considered valuable would be the values of the resistors used in the final mixer/filter portion.
Actually, yeah, the Commodore and Apple chips were pretty cool. The IIgs audio chip was pretty impressive for the 80’s… It was 32 channels athough only 15 with full ability. The annoying ‘click’ glitch was never solved, right? (I think it was NOT a limitation of the main CPU, but internal to the way the audio chip worked!)
It’s been years since I even looked at a Commodore programmer’s manual.
Hi, can anyone tell me who holds the rights for the images shown (especially the third from above)? I’m looking for a hi-res picture of an integrated circuit for the publication of my dissertation on the history word processing.
Quoting from above: “Credit for the chips go to Tim McNerney. Tim is an expert on the 4004 who has built an interactive exhibit of the chip for the Intel museum. For more information please visit the Intel 4004 35th anniversary project web site.”
Is there any way to download the two large images without the watermarks? Thank you!
Ask Tim; he holds the right to the pictures: http://alumni.media.mit.edu/~mcnerney/