Chinese semiconductor industry
- Thread starter Hendrik_2000
- Start date
- Status
Typical BIS QA:
Concerned People: Highlight serious, rational and totally valid concerns about overreaching rules that are going to affect the income of their own companies.
BIS: We disagree, followed by the most absolutely nothing lawyer jargon ridden answer ever.
Example:
Topic 38: A commenter noted that there does not appear to be a national security basis for excluding equipment sales to NAND memory fabricating facilities in China because NAND memory is so widely available on the commercial market. This regulation will harm U.S. companies and jobs while boosting the market share gain of our allies where the majority of NAND memory is manufactured. BIS
response: BIS disagrees with this commenter’s characterization of the controls. The end use control under § 744.23 and the “U.S. persons” control under § 744.6 both now reference the newly defined term “advanced-node integrated circuits” added by the SME IFR. That term specifies NAND memory as part of the criteria as well as the level of NAND memory that is a concern (i.e., NOT AND (NAND) memory integrated circuits with 128 layers or more). This higher threshold for NAND memory was intended to distinguish between the type of items easily 35 obtained on the open market and the types of NAND memory that represent national security and foreign policy concerns under the October 7 IFR.
How? Since when the number of layers of NAND chips become a NS problem? What is going to happen when more NAND with more layers than 128 becomes more common in the market in the coming years? Is BIS going to soften then? Who wrote this answer? I think I know, I can smell the Genders Studies Degree of whoever wrote this crap all the way to here.
Concerned People: Highlight serious, rational and totally valid concerns about overreaching rules that are going to affect the income of their own companies.
BIS: We disagree, followed by the most absolutely nothing lawyer jargon ridden answer ever.
Example:
Topic 38: A commenter noted that there does not appear to be a national security basis for excluding equipment sales to NAND memory fabricating facilities in China because NAND memory is so widely available on the commercial market. This regulation will harm U.S. companies and jobs while boosting the market share gain of our allies where the majority of NAND memory is manufactured. BIS
response: BIS disagrees with this commenter’s characterization of the controls. The end use control under § 744.23 and the “U.S. persons” control under § 744.6 both now reference the newly defined term “advanced-node integrated circuits” added by the SME IFR. That term specifies NAND memory as part of the criteria as well as the level of NAND memory that is a concern (i.e., NOT AND (NAND) memory integrated circuits with 128 layers or more). This higher threshold for NAND memory was intended to distinguish between the type of items easily 35 obtained on the open market and the types of NAND memory that represent national security and foreign policy concerns under the October 7 IFR.
How? Since when the number of layers of NAND chips become a NS problem? What is going to happen when more NAND with more layers than 128 becomes more common in the market in the coming years? Is BIS going to soften then? Who wrote this answer? I think I know, I can smell the Genders Studies Degree of whoever wrote this crap all the way to here.
It also added to the list of equipment restricted from going to that country to include some deep ultraviolet (DUV) lithography systems, going beyond recent Dutch regulations to keep the Netherlands' ASML from sending older DUV models and spare parts to some advanced Chinese chip factories, confirming another Reuters report.
The problem is what tphuang said in this very thread a couple of pages ago, as he is always on top of things.
What Huawei had a few years ago, would be subject to a ban if Huawei was an American company, lol.
What Huawei has today, and in production inside data centers inside China and worldwide, would be banned if Huawei was an American company, lol.
Actually, I am very angry to read such news. We are constantly bombarded with lies from the US government. They were building up this new round of bans. Then it comes out like this?!
Huawei is being turned into a monopoly.
This is the official text (for people inclined to read long legalese stuff):
The summary of the changes is at page 5
I'd wait for a deep review form expert people....
From what we can read as of today, it does not seem such a "big" thing, and nothing compared to the last year's one.
- Yes, Biren and Moore Thread are now on entity list, but this was very well expected (actually it was a miracle they were not already there).
- Yes, Nvidia H800 should be further crippled to below 300 TOPS, but is this really such a heavy blow?
- Yes, there will be further limitations on SME equipment, but above 14/16nm US firms are already out of the market the facto, and below 16nm they were already banned since last year.
Overall, I'd wait to see if there will be some last minute "surprise" that sneaks in before final version, but current version seems very AI focused and even in that scope, is far from clear the real impact on NVIDIA actual sells on the Chinese market.
The real big, heavy blows like: ASML banning, EDA banning, TSMC / Samsung foundry services banning, even not to mention the brain-dead idea of banning RISC-V and other open source stuff for China, all this seems out of the scope of current version.
EDIT: It seems ASMLis somehow affected:
Last edited:
Can someone be in the shoe of the US and explain to me why trying to ban "advanced" AI chips which are basically glorified GPU's going to prevent China from developing AI and be competitive with US AI models?
1. Machine learning with Neural Networks is by far one of the most lazy parallel tasks in the field of Computing. Which means it can be infinitely parallelized. Which means instead of using 1 fast chip you can just run 10 or 100 slower chip to do the same thing. It might require more power or cost more due to the whole extra infrastracture required. But is making Machine Learning Training more costly that much a of a big deal?
2. Even a slower chip from Nvdia should be quite useful for China to train AI models
3. What is preventing China from using the so called game console GPU's which are not banned to run the same code. Game GPU's are extremely fast as well and can run the same code using CUDA. Maybe less efficient but very viable workaround.
4. Machine Learning training is basically a one time task. Once you train your model, you can run the model for future prediction tasks with a much smaller chip or network of chips. A trained model that might need 1000 GPU's to train will be able to run on a phone GPU. So, the amount of "Advanced" Chips china needs could be much smaller than consumers need to run say AI models for Self-Driving Cars or Autonomous Drones
5. What is preventing Citizens of Friends of US who are not banned like India or Vietnam from buying up thousands of chips. putting them in suitcases or small boxes and shipping to China? Chips are made by the millions and circulated all over the world. A smuggling network will develop in a heartbeat
6. Why not transfer your data to Vietnam, train your model, transfer the trained model back to China? No need for a Chinese company to do this. Just setup a Vietnamese company in name only, with a Vietnamese owner paid by you behind everyone's back and essentially run the company from behind.
7. Why not optimize Machine Learning models using model optimization techniques which significantly reduce the size of your model without degrading performance too much. Then you can run the "light" model on a phone GPU
1. Machine learning with Neural Networks is by far one of the most lazy parallel tasks in the field of Computing. Which means it can be infinitely parallelized. Which means instead of using 1 fast chip you can just run 10 or 100 slower chip to do the same thing. It might require more power or cost more due to the whole extra infrastracture required. But is making Machine Learning Training more costly that much a of a big deal?
2. Even a slower chip from Nvdia should be quite useful for China to train AI models
3. What is preventing China from using the so called game console GPU's which are not banned to run the same code. Game GPU's are extremely fast as well and can run the same code using CUDA. Maybe less efficient but very viable workaround.
4. Machine Learning training is basically a one time task. Once you train your model, you can run the model for future prediction tasks with a much smaller chip or network of chips. A trained model that might need 1000 GPU's to train will be able to run on a phone GPU. So, the amount of "Advanced" Chips china needs could be much smaller than consumers need to run say AI models for Self-Driving Cars or Autonomous Drones
5. What is preventing Citizens of Friends of US who are not banned like India or Vietnam from buying up thousands of chips. putting them in suitcases or small boxes and shipping to China? Chips are made by the millions and circulated all over the world. A smuggling network will develop in a heartbeat
6. Why not transfer your data to Vietnam, train your model, transfer the trained model back to China? No need for a Chinese company to do this. Just setup a Vietnamese company in name only, with a Vietnamese owner paid by you behind everyone's back and essentially run the company from behind.
7. Why not optimize Machine Learning models using model optimization techniques which significantly reduce the size of your model without degrading performance too much. Then you can run the "light" model on a phone GPU
manqiangrexue
Brigadier
It's not meant to stop China. After Huawei's new phone, I think it's apparent to them that they lost that gamble, just like every other gamble they made of China failing to make something. It was done for votes to stay in office by acting like they're tough on China. To the American sheeple, banning something means you're tough, like a judge imposing a punishment. So they if they're down to nothing but Panda Express, they'll ban them from selling to China just for the nostalgic way the announcment makes Americans feel.
My gut instinct is elections are coming up in a year and Democrats have to look tough on China. A talking point for Biden would be that he destroyed Chinese progress in high tech for decades, something that Trump didn't even do.
You can directly infer US establishment's confidence of their actions on how jubilant their US media report US Gov's actions
The first big sanctions on Huawei and then the so called October surprise, were widely circulated in US media with much happiness, eagerness, and anticipation for China collapsing. You had hundreds of Twitter threads by "think-tankers" in glee, Reddit posts which had like 50k likes and some hundred awards each etc. You had US publications FT, NYTimes, Washington Post, WSJ etc all basically declaring US victory prematurely.
So coming to today's action, I searched for China on Reddit, only a single post of 200 something upvotes and 200 something comments (majority of them just saying its pointless) lol. I checked Twitter, basically no discussion at all, and the few threads that they are from some think tankers there is much gloomy feeling and a sense of helplessness. I then checked US media, not very publicised, kinda low-key while trying to keep some of the facade of victory being just around the corner even though they always reluctantly mention that China is beating the sanctions
Gone is the optimism, gone is the feeling of victory. Its all gone. This to me shows how the US establishment feels towards the semiconductor war and the now new sanctions they initiated against China.
So to conclude, its as he says it here:
Last edited:
That's a good start, but justice won't be done until they are completely and utterly ruined beyond any hope of salvation. Both materially and in spirit.
- Status