Chinese semiconductor thread II

tokenanalyst

Brigadier
Registered Member

Total investment of 1.6 billion! Inner Mongolia plans to build a high-purity quartz and products project​


Recently, the Hohhot Ecological Environment Bureau issued the " Notice on the Acceptance of Environmental Impact Assessment Documents for the Annual Production of 60,000 Tons of High-Purity Quartz Sand Semiconductor Project of Inner Mongolia Qianhe New Materials Technology Co., Ltd. " It is reported that the 60,000-ton annual high-purity quartz sand semiconductor project of Inner Mongolia Qianhe New Materials Technology Co., Ltd. is located in the Naobao Logistics Park in Xincheng District, Hohhot. The project is built by Inner Mongolia Qianhe New Materials Technology Co., Ltd. with a total investment of 1.6 billion yuan. After reaching full production, the project will form an annual production scale of 60,000 tons of high-purity quartz sand, 25,500 tons of quartz crucibles, 11,250 tons of quartz rings, and 10,000 tons of quartz tubes. The quartz raw material used in the project is imported quartz ore, and the high-purity quartz sand produced is used as a raw material for semiconductor production or directly sold according to market conditions.​
1752599227395.png

Please, Log in or Register to view URLs content!
 

tamsen_ikard

Senior Member
Registered Member
IMHO, Chinese companies should try to wean off of it as fast as they can rather than buying more of it. The only reason they should be buying more Nvidia chips is only if Huawei has a supply/capacity issue. Huawei already has chips that perform better than H20 and the issues are more to do with software ecosystem and support from companies, developers and researchers. Most of the Chinese researchers that are not affiliated with Huawei are still focused on finding innovative ways of squeezing more performance out of downgraded Nvidia chips rather than trying to do the same for Huawei or other domestic chips. That needs to change.
Without EUV Huawei will get more and more behind without any hope of catching up. Everything depends on getting the EUV as fast as possible.
 

gelgoog

Lieutenant General
Registered Member
It is for logic. Logic needs EUV. But like I keep saying AI inference can be done with analog computing. For whatever reason China does not seem to be pursuing this yet despite its obvious advantages. I keep posting the link to the Mythic AI chip but there are other competitors in the space as well.
 

latenlazy

Brigadier
It is for logic. Logic needs EUV. But like I keep saying AI inference can be done with analog computing. For whatever reason China does not seem to be pursuing this yet despite its obvious advantages. I keep posting the link to the Mythic AI chip but there are other competitors in the space as well.
Because you have to build firmware level code stacks to employ analog and this approach doesn’t scale well commercially unless you can develop some kind of standard platform. There’s work being done here across the industry (not just in China) but no one knows whether it will translate into common adoption. Honestly much more likely that China gets to sub 5 nm and Huawei gets to parity with Nvidia first.
 

tokenanalyst

Brigadier
Registered Member
It is for logic. Logic needs EUV. But like I keep saying AI inference can be done with analog computing. For whatever reason China does not seem to be pursuing this yet despite its obvious advantages. I keep posting the link to the Mythic AI chip but there are other competitors in the space as well.
The techniques that can be use to solve the issues with EUV lithography blur can be used to extend the previous generation of patterning at a decent price. Looks like that patterning shaping techniques are going to be the main focus for the next gen semiconductor devices instead of relying on lithography shear power for creating smaller devices, I think NAURA and AMEC are going to become formidable forces. For the same reason I think maybe their goal is to achieve HNA EUV as soon as possible instead of LNA EUV given that they probably can continue advancing with other techniques.
 

LanceD23

Junior Member
Registered Member
Is important but not absolutely necessary. EUV is not the savior that the industry thought.
Huawei 910D based on 7nm chiplet is doing fine. Market share for Huawei AI chip has risen to 37%. 910D is sufficient for now.

Jansen huang of Nvidia is drooling to go back to China with H20. He is trying to hook up with Lei Jun of Xiaomi at beijing. Xiaomi plays the role of foreign tech collaborator or traitor much like Lenovo did before.

There are.alot of noises they are doing EUV trial run at shenzhen.
 
Last edited:
Top