Chinese semiconductor industry

Status
Not open for further replies.
B

BlackWindMnt

Captain
Registered Member
horse said:
Yeah, exactly!

There lies the big question mark.

1) Why do this ban of that H100 chip now? This chip is developed in China, close to full completion. Now it is banned from China? Are those guys serious? Are those guy retarded?

2) Although I have not looked that website that ranked servers lately, the best servers in the world and by a lot where no one is going to catch up soon, are Chinese made. This H100 will not make a difference in the market for data center products. Why do we think Huawei was able to make so many deals across ASEAN, the Middle East, Africa, and South America, because they were providing a completion solution, the network, the data center, and any further products to do whatever you want to do, even industrial fish farming.

The people in the US government dreaming up these bans, seem unaware of the technology world at large.

:oops:
Click to expand...
Just look at the west and the simple relationship of Russia being a energy and food provider and Europe being an energy and food consumer is something they weren't able to foresee what would happen when implementing the sanctions. That is from a view the western elites actually give a tiny bit of shit about their people. See if they don't give a shit about their people then yeah its actually a well though out plan so western big capital can buy up even more property.

Now take something multiple orders more complex like semi conductor supply lines im sure they only have a very small picture and have no clue how the big picture looks like.
 
tphuang

tphuang

Lieutenant General
Staff member
Super Moderator
VIP Professional
Registered Member
FairAndUnbiased said:
big picture: they're paying high salaries to train Chinese GPU design engineers.

where do they go?

Please, Log in or Register to view URLs content!
Click to expand...
Awesome, now ban Nvidia from further development in China. That would be a fair reciprocal move. Chinese AI/cloud service companies were given no time to make adjustment. Neither should Nvidia be given time to further research on Chinese brains. Then, all those Nvidia employees can go work at Biren Technology and Nvidia has to take longer to finish its new chips.

In the meanwhile, you need to work on enticing more ethnic Chinese engineers in this field to move back to China.

So, I took look at what Alibaba cloud, which has like 1/3 market share in China uses. Looks like a lot of their ECSs are using A100 GPUs. I know they had developed Yitian-710, but it looks like it is still in low production and is not an AI accelerator.
Please, Log in or Register to view URLs content!
Still, this seems like a good move toward using domestic product on domestic cloud services. It also seems to me that ramp up for production even in a large tech firm like Alibaba could take a while. It took them 6 months after announcement to even get to this point.

Then, I took a look at Baidu cloud and what they had domestically. That was the Kunlun 2 chip
Please, Log in or Register to view URLs content!
In this case, it does seem like a chip designed for AI. However, it seems designed just for Baidu Brain 7.0. I'm not sure if this will ever appear outside of Baidu data centers anytime soon. On top of that, there hasn't been news about it even in their recent earnings call. Also, the good people at XPeng that is building a large self driving center with Baidu is still using A100 based on their recent social media post
Please, Log in or Register to view URLs content!
Well, that's discouraging. That tells me Kunlun 2 at this point is probably still just testing out in limited number of Baidu server boxes.

Then I took a look at Huawei/Hisilicon. Looks like they actually had a really compelling product in Ascend 910 when it got launched. It was actually using 7 nm process and had better performance than the previous generation V100 (which was the flagship product of Nvidia back in 2019 and using 12 nm proces)
Please, Log in or Register to view URLs content!
But as we know, Huawei/Hisilicon got crushed by sanctions. Now, I'm optimistic that Huawei will be back and designing great chips with the domestic line. However, they will have to balance the need for AI chips with smartphone chips and other areas. But this is at least a good contender.

Which brings us to Biren BR100. I read a great article here. Honestly, they seem to have some really solid people behind the project and a lot of funding. And more importantly, they have already started shipping their flagship BR100/104 product.
Please, Log in or Register to view URLs content!
Looks like it won one of the awards here along with Huawei cloud's drug design platform and Horizon Robotics's Journey 5 chips.
Please, Log in or Register to view URLs content!
Biren has really moved fast. They managed to launch this flagship chip in 3 years. It was designed from start to be 10 times the speed of the leading edge V100 chip at the time. So when it got released, it is best in the world. Although probably behind H100 when the latter comes out. So, China slowing down Nvidia development by not allowing them to finish their H100 project in China is probably fair.
Just 5 months ago, they said they had successfully tested BR100. Now, they are able to already ship it. That's impressive.
Please, Log in or Register to view URLs content!
Again, Biren believes their product is even competitive vs H100. It aimed very high in the design phase.
Please, Log in or Register to view URLs content!

Looks like their initial customer is probably going to be Alibaba cloud, but I'm sure more people will join later. If their project was supported by Alibaba cloud, then it makes sense they would've been working with Alibaba to fine tune their GPU up until now. If the performance is as good as advertised, it's just a matter of ramping up production. They'd be well served now to stock enough TSMC chips for a while until they can get a similar chip designed with SMIC 7nm process. TSMC is seeing orders slow down, so Biren should put in a major order in for the next year.

I'm also curious about this other company enflame mentioned here on GT article.
Please, Log in or Register to view URLs content!
 
T

tokenanalyst

Brigadier
Registered Member
T

tokenanalyst

Brigadier
Registered Member

To think that one day we will have rovers navigating the hellish surface of Venus​


Demonstration of 4H-SiC CMOS digital IC gates based on the mainstream 6-inch wafer processing technique.​

Tongtong Yang1, Yan Wang1, 2, and Ruifeng Yue1, 2
  • 1. School of Integrated Circuits, Tsinghua University, Beijing 100084, China
  • 2. Beijing National Research Center for Information Science and Technology, Beijing 100084, China

Although silicon carbide (SiC) power devices have been proven to be stable and reliable for high temperature operation, the present control and drive circuits are based on silicon material, which restricts the maximum operating temperature of the power electronics system[1]. Developing SiC integrated circuits (ICs) is beneficial for realizing the full potential of SiC material in high temperature applications[2-6]. Compared with other technology, the complementary-metal-oxide-semiconductor (CMOS) technology could achieve full rail-to-rail output voltage switching, low power losses and temperature-independent logic levels[7, 8]. Unfortunately, the reported SiC CMOS ICs in the literature are fabricated using specially developed technology, which is not compatible with the mainstream 4H-SiC power device processing technology.

Motivated by developing 4H-SiC based power ICs which integrates the control circuits and power device in a single chip, this article reports the implementation of fundamental CMOS-based digital gates based on 6-inch SiC wafer processing technology, in which the digital gates and power devices are fabricated simultaneously. The fundamental characteristics of the inverter and NAND gates are characterized and analyzed. The experimental results that are obtained by this work will serve as useful guides for driving the development of SiC-based power ICs.


1662132679478.png
1662132734135.png

Please, Log in or Register to view URLs content!
 
A

Appix

Senior Member
Registered Member

With new China AI chip restrictions, U.S. takes aim at a critical niche​


Sept 1 (Reuters) - The United States beefed up its effort to cut off the flow of advanced technology to China by instructing Nvidia Corp
Please, Log in or Register to view URLs content!
and Advanced Micro Devices
Please, Log in or Register to view URLs content!
to stop sending their flagship artificial intelligence chips there.

While the news shocked the chip sector by the time markets closed Thursday, sending the Philadelphia semiconductor index down 1.9% and Nvidia and AMD down 7.6% and 3% respectively, the letters from the U.S. officials appeared to target a narrow but critical part of China's computing industry.

The regulations appear to focus on chips called GPUs with the most powerful computing capabilities, a critical but niche market with only two meaningful players, Nvidia and AMD. Their only potential rival - Intel Corp
Please, Log in or Register to view URLs content!
- is trying to break into the market but has not released competitive products.

Originally designed for video games, the usage of GPUs, or graphic processing units, have been expanded to a wider array of applications that include handling artificial intelligence work like image recognition, categorizing cat photos or scouring digital satellite imagery for military equipment. Because all the chip suppliers are American, the U.S. controls access to the technology.

Some national security experts saw the U.S. move as a long time coming.

GPUs "have been totally uncontrolled to China and to Russia, so in a lot of ways I see this action as kind of catching up to where the controls probably should have been if we were really serious about trying to slow China’s AI growth," said Emily Kilcrease, a senior fellow at the Center for a New American Security.

The U.S. Department of Commerce, which declined to comment on the specifics of whatever new rules it may be developing, appears to have targeted the effort narrowly.

The only products Nvidia said would be affected are its A100 and H100 chips. Those chips cost tens of thousands of dollars each, with full computers containing the chips costing hundreds of thousands of dollars.

Similarly, AMD said that only its most powerful MI250 chip - a version of which is being used at Oak Ridge National Laboratory, one of several U.S. supercomputing sites that supports nuclear weapons - is affected by the new requirement. Less powerful chips such as AMD's MI210 and below are not affected.

What the affected chips share is the ability to carry out calculations for artificial intelligence work quickly, at huge scale and with high precision. Less powerful AI chips can work quickly at lower levels of precision, which is sufficient for tagging photos of friends and where the cost of an occasional mistake is low - but are insufficient for designing fighter jets.

The only major market rival to AMD and Nvidia's chips is Intel's still-unreleased Ponte Vecchio chip, whose first customer is Argonne National Lab, another U.S. installation that supports nuclear
weapons.

"While we understand the U.S. Government is continuing to look at new restrictions, no new export control rules have been published and there are currently no changes to our business," Intel told Reuters in a statement. "We are closely monitoring the process."

Source:
Please, Log in or Register to view URLs content!
 
FairAndUnbiased

FairAndUnbiased

Brigadier
Registered Member
tokenanalyst said:

To think that one day we will have rovers navigating the hellish surface of Venus​


Demonstration of 4H-SiC CMOS digital IC gates based on the mainstream 6-inch wafer processing technique.​

Tongtong Yang1, Yan Wang1, 2, and Ruifeng Yue1, 2
  • 1. School of Integrated Circuits, Tsinghua University, Beijing 100084, China
  • 2. Beijing National Research Center for Information Science and Technology, Beijing 100084, China

Although silicon carbide (SiC) power devices have been proven to be stable and reliable for high temperature operation, the present control and drive circuits are based on silicon material, which restricts the maximum operating temperature of the power electronics system[1]. Developing SiC integrated circuits (ICs) is beneficial for realizing the full potential of SiC material in high temperature applications[2-6]. Compared with other technology, the complementary-metal-oxide-semiconductor (CMOS) technology could achieve full rail-to-rail output voltage switching, low power losses and temperature-independent logic levels[7, 8]. Unfortunately, the reported SiC CMOS ICs in the literature are fabricated using specially developed technology, which is not compatible with the mainstream 4H-SiC power device processing technology.

Motivated by developing 4H-SiC based power ICs which integrates the control circuits and power device in a single chip, this article reports the implementation of fundamental CMOS-based digital gates based on 6-inch SiC wafer processing technology, in which the digital gates and power devices are fabricated simultaneously. The fundamental characteristics of the inverter and NAND gates are characterized and analyzed. The experimental results that are obtained by this work will serve as useful guides for driving the development of SiC-based power ICs.


View attachment 96725
View attachment 96726

Please, Log in or Register to view URLs content!
Click to expand...
So this looks to me like these ICs can be run on input voltage and directly interacting with the "high voltage" side, rather than say a MCU running at TTL voltage and controlling the high voltage through a relay or MOSFET. That would in turn eliminate the requirement for an external voltage regulator.

For instance a 24V motor controller is actually a MCU controller running on say, 5V, that outputs a 5V signal to a power MOSFET gate to switch 24V. But this can make a 24V motor controller that just outputs 24V.

Did I understand correctly?
 
A

Appix

Senior Member
Registered Member
Who cared about semiconductors in China during the 1980s and 1990s. Most of us were sweatshop slaves barely making end meets and before that subsistence peasants. We did not care or understand the outside world.
 
FairAndUnbiased

FairAndUnbiased

Brigadier
Registered Member
Appix said:
Who cared about semiconductors in China during the 1980s and 1990s. Most of us were sweatshop slaves barely making end meets and before that subsistence peasants. We did not care or understand the outside world.
Click to expand...
This is a bizarre outburst. But also, Huahong NEC was established in 1997 and SMIC in 2000 (so prep work was in the 90s). China has always punched far above it's weight in tech vs GDP per capita.
 
Status
Not open for further replies.
Top