Chinese semiconductor industry

Status
Not open for further replies.

HighGround

Senior Member
Registered Member
Go ahead, I think it is a great news for China's chip development

It's not. It's actually good to have a competitor.

I really want to see an industry effort at standardising a programming model to rival CUDA instead of having AscendCL, SUPA, MUSA etc. all co-existing. You could allow for some vendor-specific APIs to account for quirky hardware features but should be de-duplicating the generic stuff as much as possible so that a developer doesn't need to learn three ways to code the same thing

This process is probably going to happen naturally at some point. No need to try to centrally coordinate it.

This sort of competition is largely a good thing.

I can see how central planners would want to "harmonize" these efforts because they see this is a "tech race" with United States, and in certain contexts this approach makes sense. But I don't think this approach is efficient or good for the industry.

You need to let the industry organically determine what is useless and what isn't, the winners of this competition will quickly lap up the spoils without government's directions.
 

henrik

Senior Member
Registered Member
well.

Please, Log in or Register to view URLs content!

Taiwan’s manufacturing output -16.4% year-on-year in May, the 12th straight negative month, with semiconductor output -25.9% and LCD panel output -22.9%, media report. From January-May, Taiwan’s semiconductor output -23.5%..

after Samsung historic loss in chip manufacturing output its time for Taiwan. thanks to sanctions and tech war initiated by the USA and China's progress in indigenous semiconductor manufacturing. IMHO this is the beginning of downfall.

Nvidia stock price is crumbling, right after new sanctions are announced.
 

tphuang

Lieutenant General
Staff member
Super Moderator
VIP Professional
Registered Member
if you are concerned about not having access to Nvidia AI chips. These are just news from today

Please, Log in or Register to view URLs content!
当天,中国联通启动了
Please, Log in or Register to view URLs content!
产业知识产权运营中心建设、成立中国联通-科大讯飞人工智能反诈联合实验室、成立中国联通昇腾人工智能创新实验室,一中心两实验室的启动和揭牌彰显了中国联通携手合作伙伴,进一步深耕网络安全和人工智能领域的决心和信心,为后续技术攻关奠定了基础。
so China Unicom is working with iFlyTek for an Anti-fraud AI lab
it's working with Ascend for an AI innovation lab
So again, China Unicom is using Ascend and Huawei tech

Please, Log in or Register to view URLs content!
Guangzhou public AI computation center is also using Huawei/Ascend tech

So, I'm not concerned at all. In fact, it would be good if Tencent can shut up and start using domestic chips.
 

theorlonator

Junior Member
Registered Member
if you are concerned about not having access to Nvidia AI chips. These are just news from today

Please, Log in or Register to view URLs content!

so China Unicom is working with iFlyTek for an Anti-fraud AI lab
it's working with Ascend for an AI innovation lab
So again, China Unicom is using Ascend and Huawei tech

Please, Log in or Register to view URLs content!
Guangzhou public AI computation center is also using Huawei/Ascend tech

So, I'm not concerned at all. In fact, it would be good if Tencent can shut up and start using domestic chips.
Are these all from stockpiling or is SMIC 7nm good enough for this much demand?
 

tokenanalyst

Brigadier
Registered Member

Wanye Enterprise: R&D investment has increased year-on-year, and the new R&D base will be completed this year.​


Wanye Enterprise's 2022 annual report shows that in the face of the company's industrial transfer, increase in new product layout, continuous update of existing products and large investment in research and development, Wanye Enterprise's operating income in 2022 will be 1.158 billion yuan. Among them, the special semiconductor integrated circuit equipment manufacturing business achieved sales revenue of 206 million yuan, a year-on-year increase of 67.96%, and the real estate business realized sales revenue of 898 million yuan, a year-on-year increase of 31.86%. At the same time, the operating income of Kaishitong, a subsidiary of the company, will be 139.8929 million yuan in 2022, a year-on-year increase of 12.00%. The company's subsidiary Jiaxin Semiconductor will have an operating income of 66.6315 million yuan in 2022, a year-on-year increase of 1188.29%.

The company accelerated the transfer of strategic focus to the semiconductor industry, and the R&D investment increased significantly year-on-year

Wanye Enterprise said that with the increase in the layout of new products and the continuous improvement of existing products, the company continues to pay attention to R&D investment, including: R&D personnel salaries, material costs related to R&D projects, engineering costs, and collaborative development costs. Therefore, the company's research and development expenses in 2022 will be 108 million yuan, a year-on-year increase of 136.99%, and the ratio of research and development expenses to operating income will be 9.34%. Wanye said that the follow-up company will make reasonable arrangements for R&D investment according to business development planning, R&D project progress, company profitability and capital situation.

Jiaxin Semiconductor Industrial Park was completed this year and has become one of the largest R&D bases for integrated circuit front-end equipment in East China

Referring to the company's development direction this year, Wanye Enterprises said that its subsidiary Kaishitong has always adhered to customer-centricity, and has established R&D and customer service centers in Shanghai Jinqiao, Shanghai Zhangjiang, Shanghai Lingang, Beijing Yizhuang and Anhui Hefei. . Among them, the Jinqiao base is the latest industrialization center, including technology research and development, CIP improvement, component assembly and verification, manufacturing, process verification, customer training and other functions. The base will greatly expand the production capacity of the existing equipment of Kaishitong ion implanter, and can provide customers and partners with the evaluation of a full range of products, shortening the time from technology verification to customer introduction.

At the same time, the 109-acre industrial park invested and built by Jiaxin Semiconductor, a subsidiary of the company, will complete the completion and acceptance in 2023, and will become one of the largest R&D and manufacturing bases for integrated circuit front-end equipment in East China. In addition, Jiaxin Semiconductor has a nationwide layout, and has set up service stations in Shanghai Pudong, Tianjin, Qingdao, and Shenzhen to meet 7×24-hour customer service.

In general, the company has promoted the continuous improvement of the comprehensive capability system from technical strength to market expansion by realizing the "spreading" of R&D, production, and customer service systems across the country, and accelerated the development of R&D capabilities to a new height, consolidating industrial clusters Synergistic advantages will open up a broader market structure for future development.

Please, Log in or Register to view URLs content!
 

huemens

Junior Member
Registered Member
I really want to see an industry effort at standardising a programming model to rival CUDA instead of having AscendCL, SUPA, MUSA etc. all co-existing. You could allow for some vendor-specific APIs to account for quirky hardware features but should be de-duplicating the generic stuff as much as possible so that a developer doesn't need to learn three ways to code the same thing
What about the already open source OpenCL. What is it's limitation compared to CUDA?
 

gelgoog

Lieutenant General
Registered Member
Lack of applications which use it basically. NVIDIA came to the market with CUDA first and spent huge amounts of money either porting existing software to CUDA, or giving GPU cards to 3rd parties to port their software to CUDA. Unless a similar effort is made for OpenCL it will never get to the same level of market penetration. If you want to code your own application, you will typically find there are dozens of CUDA libraries or frameworks already available, and only marginal OpenCL support. Which means you would need to do a lot of work yourself if you want to get anything done.
 
Last edited:
Status
Not open for further replies.
Top