| View previous topic :: View next topic |
| Author |
Message |
Imgema
Joined: 07 Dec 2014 Posts: 5
|
Posted: Sun Dec 07, 2014 8:50 am Post subject: The best, highest resolution die scans so far? |
|
|
These are the ones i have found, so far
80286 - http://visual6502.org/images/80286/Intel_80286_die_shot_20x_1a_9500w.jpg
And this one i found in this forum: http://www.g0b.fr/cpu/Test.htm
This second chip, i have no idea how many transistors it has so i don't know how much more advanced than the 80286 is.
My question is, are there any CPUs, more modern/advanced than these, where we can see the transistors at the same clarity? I'm not interested for older CPUs with less transistors, there are already many detailed scans for them. They don't need such high resolution or such powerful microscopes for their transistors to be visible.
The 80286 has 130.000+ transistors and a 9k x 9k scan can reveal the details but how about a 80386 or a 80486? I've seen a few good shots of the 386 die they are far too blurry compared to the 80286 shot above and any CPU die from 80486 and above (1 million transistors and counting) just isn't big enough to reveal such detail.
I know i'm asking a lot, a detailed 80486 shot would probably need a huge 1GB picture, but is it possible? How about a pentium? (and to think that modern CPUs exceed 2 billion transistors... i can't even begin to imagine how can someone design such a die) |
|
| Back to top |
|
 |
frag_
Joined: 17 Nov 2008 Posts: 4015 Location: Estonia
|
Posted: Sun Dec 07, 2014 7:44 pm Post subject: |
|
|
The most complex cpu with transistor-level photo I am aware is PlayStation 1 CPU.
There is reverse engineering project on it, complete image is ~50k x 50k pixels.
http://psxdev.ru/download
Unfortunately most of this cpu is based on standard cells,
not very interesting, but easier to reverse than full custom design. |
|
| Back to top |
|
 |
Imgema
Joined: 07 Dec 2014 Posts: 5
|
Posted: Sun Dec 07, 2014 8:42 pm Post subject: |
|
|
Nice find, thanks!  |
|
| Back to top |
|
 |
g0b

Joined: 18 Jul 2006 Posts: 1385 Location: France
|
Posted: Mon Dec 08, 2014 5:22 am Post subject: Re: The best, highest resolution die scans so far? |
|
|
| Imgema wrote: |
This second chip, i have no idea how many transistors it has so i don't know how much more advanced than the 80286 is.
|
VR10000 has circuit size roughly 4x thinner than 286 , and 50x more transistors.
486 is 1 micron circuit size, not far from 286.... So it should not be difficult to make a picture at transistor level... _________________ Life is a long lesson in humility |
|
| Back to top |
|
 |
Imgema
Joined: 07 Dec 2014 Posts: 5
|
Posted: Mon Dec 08, 2014 2:16 pm Post subject: Re: The best, highest resolution die scans so far? |
|
|
| g0b wrote: | | 486 is 1 micron circuit size, not far from 286.... So it should not be difficult to make a picture at transistor level... |
Οh, i thought there was a huge difference because the 286 has 134.000 transistors while the 486 has more than a million. So i guess the die is bigger and the transistors more dense? |
|
| Back to top |
|
 |
g0b

Joined: 18 Jul 2006 Posts: 1385 Location: France
|
Posted: Sat Dec 13, 2014 3:22 am Post subject: |
|
|
Anyway, the maximum resolution that can be achieved with an optical microscope (visible light) is around 0.2 micron (200 nm) so don't expect to see transistors on a P3+ die... _________________ Life is a long lesson in humility |
|
| Back to top |
|
 |
Imgema
Joined: 07 Dec 2014 Posts: 5
|
Posted: Sat Dec 13, 2014 3:11 pm Post subject: |
|
|
Well, if the maximum possible is a Pentium 2, i will be happy  |
|
| Back to top |
|
 |
CPUShack

Joined: 16 Jun 2003 Posts: 34259 Location: State of Jefferson, USA
|
|
| Back to top |
|
 |
Birdman.

Joined: 13 Nov 2012 Posts: 833 Location: Finland
|
Posted: Sat Dec 13, 2014 5:59 pm Post subject: |
|
|
Unfortunately, all DEC Alphas I've opened so far have a thick coating on their dies that only allows them reflect yellowish colors. I'll try to release my yellowish die shots as soon as I can although they don't show the structures as well as expected with only small color differences. I haven't tested yet whether that coating limits sharpness at high magnification, but they may look messy.
I think that R4400 is a very good die for high-res picture, and for example Am29050, P5 Pentium, PA-7xxx, SuperSPARC, MC68040 and TX486DLC should also look great at very high resolution. Later CPUs usually have standard cell areas and more metal layers that cover many details. |
|
| Back to top |
|
 |
frag_
Joined: 17 Nov 2008 Posts: 4015 Location: Estonia
|
Posted: Sat Dec 13, 2014 6:06 pm Post subject: |
|
|
I have STM (Scanning Tunneling Microscope) at work
with 0.2-1.0 nm resolution (depending on the material),
it's potential winner here
But it's overkill for old dies and I don't know how to deal with modern ones. |
|
| Back to top |
|
 |
|