3 ms·

You say that as if it didn't just turn out that China has been producing 7nm chips for at least a year.

onepointsixC 4y ago
Real Question is what yields have they been producing those chips at. Mining chips pay for themselves, so it's not as much of a big deal if your process is bad and only 20% of your wafer is usable. If you want to make a super computer out of chips from a process with 20% yields, then you're spending absurd amounts.
- sudosysgen 4y ago
  I have a question for you. Do you know that, or why, China stopped submitting new entries to the Top500, while numerous sources are claiming China brought online exascale computers with custom architectures? Mining chips don't just simply pay for themselves. If it was cheaper to get the chips fabbed at Samsung or TSMC at any other lithography they'd be fabbed there. The company that made the miners isn't even Chinese. That it was fabbed there means that not only was there yield leftover from whatever state project was required, it was also commercially competitive.
  - kilpikaarna 4y ago
    This sounds interesting. Source?
nomel 4y ago
I think it's generally accepted that this was a copy paste of the TSMC process. I think we'll have to wait to see if they can advance this, on their own. As the other person pointed out, a good indicator would be yield.
- sudosysgen 4y ago
  It's a copy past of the TSMC process because they were able to offer far higher higher salaries and attract TSMC talent. Given the reality that they have a lot of that talent and more each year and given the excellent track record of SMIC leadership, they will probably be able to advance on this, limited by the lithography machines they can procure. If their yield was abysmal, no one would be willing to buy the chips since they aren't selling them at a loss and have plenty of public sector demand. The fact that they were able to sell them commercially and thus competitively means that their yields are okay at least.
  - nomel 4y ago
    > limited by the lithography machines they can procure I would consider that a profound hurdle to advancement. Nvidia has already moved to 5nm (released later this year).
    - sudosysgen 4y ago
      It is a hurdle, but as new processes slow it is not insurmountable to raw power.
- adrian_b 4y ago
  Only people without experience in semiconductor manufacturing imagine that it is possible to copy and paste a CMOS manufacturing process. Even with perfect knowledge about how a process is done at TSMC, implementing it in another fab with different equipment requires a very large amount of work done by very competent engineers, to substitute some process steps and tweak the technological parameters for the others. Someone who succeeds to reproduce a stolen manufacturing process is very likely able to further develop such a process on their own. Even within a single company, i.e. when having all the technical documentation and when having continuous access to all the people working there, it is extremely challenging to build a second semiconductor plant that will be able to make the same chips as a previous plant, with the same manufacturing process. It took Intel many years of huge efforts until they have perfected a methodology to "copy exactly", to be able to build additional fabs for expanding the production of an already existing processor.
  - nomel 4y ago
    > very large amount of work done by very competent engineers, Of course, and this is partly why they were successful: https://www.bloomberg.com/news/articles/2022-07-27/synopsys-backed-china-chip-firm-accused-of-poaching-tsmc-talent https://www.bloomberg.com/news/articles/2022-07-27/synopsys-... > Someone who succeeds to reproduce a stolen manufacturing process is very likely able to further develop such a process on their own. And, this is why we'll have to wait and see if their current, successful, strategy will result in future, independent, advances.

Real Question is what yields have they been producing those chips at. Mining chips pay for themselves, so it's not as much of a big deal if your process is bad and only 20% of your wafer is usable. If you want to make a super computer out of chips from a process with 20% yields, then you're spending absurd amounts.

sudosysgen 4y ago
I have a question for you. Do you know that, or why, China stopped submitting new entries to the Top500, while numerous sources are claiming China brought online exascale computers with custom architectures? Mining chips don't just simply pay for themselves. If it was cheaper to get the chips fabbed at Samsung or TSMC at any other lithography they'd be fabbed there. The company that made the miners isn't even Chinese. That it was fabbed there means that not only was there yield leftover from whatever state project was required, it was also commercially competitive.
- kilpikaarna 4y ago
  This sounds interesting. Source?

nomel 4y ago

I think it's generally accepted that this was a copy paste of the TSMC process. I think we'll have to wait to see if they can advance this, on their own. As the other person pointed out, a good indicator would be yield.

sudosysgen 4y ago
It's a copy past of the TSMC process because they were able to offer far higher higher salaries and attract TSMC talent. Given the reality that they have a lot of that talent and more each year and given the excellent track record of SMIC leadership, they will probably be able to advance on this, limited by the lithography machines they can procure. If their yield was abysmal, no one would be willing to buy the chips since they aren't selling them at a loss and have plenty of public sector demand. The fact that they were able to sell them commercially and thus competitively means that their yields are okay at least.
- nomel 4y ago
  > limited by the lithography machines they can procure I would consider that a profound hurdle to advancement. Nvidia has already moved to 5nm (released later this year).
  - sudosysgen 4y ago
    It is a hurdle, but as new processes slow it is not insurmountable to raw power.
adrian_b 4y ago
Only people without experience in semiconductor manufacturing imagine that it is possible to copy and paste a CMOS manufacturing process. Even with perfect knowledge about how a process is done at TSMC, implementing it in another fab with different equipment requires a very large amount of work done by very competent engineers, to substitute some process steps and tweak the technological parameters for the others. Someone who succeeds to reproduce a stolen manufacturing process is very likely able to further develop such a process on their own. Even within a single company, i.e. when having all the technical documentation and when having continuous access to all the people working there, it is extremely challenging to build a second semiconductor plant that will be able to make the same chips as a previous plant, with the same manufacturing process. It took Intel many years of huge efforts until they have perfected a methodology to "copy exactly", to be able to build additional fabs for expanding the production of an already existing processor.
- nomel 4y ago
  > very large amount of work done by very competent engineers, Of course, and this is partly why they were successful: https://www.bloomberg.com/news/articles/2022-07-27/synopsys-backed-china-chip-firm-accused-of-poaching-tsmc-talent https://www.bloomberg.com/news/articles/2022-07-27/synopsys-... > Someone who succeeds to reproduce a stolen manufacturing process is very likely able to further develop such a process on their own. And, this is why we'll have to wait and see if their current, successful, strategy will result in future, independent, advances.