In recent years I've become more and more worried that work on RISC-V is just a way of empowering non-democratic governments the world over by giving them an escape hatch out of western controlled IPs. Yes it's better that these things are not controlled by big companies and yes this should improve innovation in the hardware space but what are we trading off when we get that benefit?
They already did designs based on Alpha and MIPS ISAs. That genie definitely isn’t going back into the bottle.
Instead, we’ve moved to limit their ability to manufacture. All the chip plans in the world are worthless if you can’t fabricate them. ASML was largely funded with US DoD money and they seem to have veto power when it comes to exports.
Russia’s cutting edge tech is something like 180nm with promised for 90/65nm still being pipe dreams.
China has managed small amounts of 32nm and talks about 22/14nm stuff, but they can’t get new equipment and all of that is mostly gleaned from the US outsourcing so much.
This puts these countries 10-20 years behind and that’s if we don’t consider the nodes that are essentially complete, but waiting on the resources to build them out.
SMIC's 14nm has been in full production for years now at this point, and is rumored to have TSMC equivalent yields. The Kirin 710A is one example of a production chip made on that node.
China has been effectively cut off from EUV steppers, but has used that fact to put a lot of money into developing their own by any means necessary. They have a pretty decent track record of developing manufacturing experience in something when they get cut off from importing it.
SMIC 14nm transistor density sits about halfway between Intel 22nm and Intel 14nm. Intels SRAM density was almost twice that of other 14nm processes putting SMIC much closer to Intel 22nm. I’d note Intel’s 22nm is 11 years old and their 14nm is 8 years old.
Last I checked, SMIC still wasn’t ramping up 14nm production until later this year.
I believe this was directly impacted to problems getting more equipment. High yields on not much equipment also doesn’t matter too much. If that equipment was easy to design and make, then everyone would be doing it. Even a year or two makes a huge difference in this market.
As a side note though, nodes have very little impact on weapons themselves. Most missiles and similar weapons use positively huge nodes because they are more reliable and less susceptible to interference.
We could land a rocket on the moon with 350 times less compute power than your highschool TI calculator possessed. A handful of 180nm chips are more than up to the task. Even ancient 600nm Pentium designs would be more than enough.
People tend to think about weapons, but the real consequences are in the design and implementation stages.
That's not a bad thing by any means. It's making the best of a bad situation (non-democratic governments that are, to varying extents, anti-West) by incenting those governments to cooperate with the West on open alternatives to proprietary IP. Since such cooperation would basically never happen if it wasn't for the open IP-alternatives being available (they would develop proprietary IP of their own instead, as we've seen so many times already!) I'd call that a pretty good deal!
Sounds like a soft vulnerable world hypothesis. I think it's training, collaboration, motivation, and funding of engineers that really matters though. I think what's happening in Ukraine shows that as long as brain drain happens it doesn't matter if knowledge is open because it cannot be used effectively.
The conflation of ideology and tech as shown in this post is even more reasons to have indigenous and maybe open IP to prevent "west" blockade. The non-democratic Chinese dont give a shit about the ideology of their customer. I prefer tech and commerce be that way instead of an avenue of virtue signalling.
