It's logical to assume China (Huawei,smic etc) will announce something big, like the opening of a "domestic" semi line, before the end of this year: to gain the momentum as well as posturing to biden - Trump will lose w 99% certainty lol - that should biden continue w trump's semicon war china is and will be ready.. or that US semicon co's will start losing huge sales in china.
Chinese semiconductor industry
- Thread starter Hendrik_2000
- Start date
- Status
Hi machupicu,
Can't wait, I'm really looking forward to it, One by one all the topic we discuss will be reveal.
I doubt so. In eyes of China, Trump bring more profit than biden. That's why China doesn't response aggressive against US at this moment.
So far from news available SMIC obeyed US instructions, that's mean they stop supplying Huawei. I'm not optimistic SMIC can overcomes US sanctions in short of period time (if that's happen). Its not like SMIC can build domestic substituted in shorts period of time (without patent infringement). Bear in mind SMIC is public listed, so what they done need to disclose on public so shareholder can make decisions whether to buy, hold or sell share. All their movement is under public radar.
Is much more expensive to produce product when patent barrier already established. Consider many existing SMEE current product based on 90nm and below(officially), many patent on that's technology already expired. Its easy to SMEE to produce. when comes to 45nm and above hurdle is much higher. Many patent already filed on that's particular technology.
In laymen term, if EUV ASML need 3 step process(efficient) SMEE need to build more step (inefficient) in order to bypass patent barrier. When that's happened, the cost of foundry become higher, if price war happened, the less efficient foundry become loser. well if government subsidies it will slap by anti dumping by other country.
they only way out, Is china subverted entire Semiconductor industry. define new standard, build new architecture, design new language from machine code to high level programming language. Many people doesn't realize, US tech input is higher than people imagine.
well Huawei is sanction because they touch bottom line US tech dominant. Huawei want to replace current network architecture(not 5G) is redline for US. They want to build new IP essentially redefine how internet work in the future.
China doesn't want to escalate yet, knowing trump will be voted out.
On smic, of course it will abide by the BIS rules, but nothing to prevent it from establishing a new line with smee machines and other non-US equipment.
We all are waiting for that 28nm res machine from smee.
On new standards for internet, there is a body for that, ICANN(?), it will take time and lots of discussions first, the US will find alliances to counter China's side.
Based on 14nm node FABS by Samsung, so another potential customer for SMIC.
from cnTechPost
› › Baidu exec says chips developed by the company have been deployed on its search engine
Baidu exec says chips developed by the company have been deployed on its search engine
2020-10-14 19:30:25 GMT+8 | cnTechPost
0
Baidu's cloud-based universal AI chip Baidu Kunlun, developed by Baidu, is already in use on Baidu's own search engine, which is handling a large number of requests on searches every day, Baidu vice president Wu Tian said on October 14.
She said the chip brings the new semantic understanding technology of artificial intelligence to search to improve the effectiveness of search.
Wu Tian delivered a keynote speech at the Digital Economy Leaders Forum at the opening of the 2020 China International Infocomm Exhibition on Oct. 14.
In addition to its own business, Baidu is also opening up its AI capabilities to the outside world to support the various needs of enterprises when it comes to intelligence, such as in the energy industry, where safety production, intelligent dispatching, inspection, and intelligent customer service are all very important application scenarios, she said.
However, after the infrastructure has been provided for the enterprise, the talent is needed to make better use of this infrastructure.
She believes that in the process of the digital upgrade of enterprises, the most lacking is actually a composite talent, which understands both the boundaries of technology and business understanding.
Based on the infrastructure provided by Baidu, has also been committed to building a training system to be able to promote the training of talent in the whole society, said Wu Tian, adding that currently Baidu has trained more than 1 million AI talents, and in the future hope to train more.
from cnTechPost
› › Baidu exec says chips developed by the company have been deployed on its search engine
Baidu exec says chips developed by the company have been deployed on its search engine
2020-10-14 19:30:25 GMT+8 | cnTechPost
0
Baidu's cloud-based universal AI chip Baidu Kunlun, developed by Baidu, is already in use on Baidu's own search engine, which is handling a large number of requests on searches every day, Baidu vice president Wu Tian said on October 14.
She said the chip brings the new semantic understanding technology of artificial intelligence to search to improve the effectiveness of search.
Wu Tian delivered a keynote speech at the Digital Economy Leaders Forum at the opening of the 2020 China International Infocomm Exhibition on Oct. 14.
In addition to its own business, Baidu is also opening up its AI capabilities to the outside world to support the various needs of enterprises when it comes to intelligence, such as in the energy industry, where safety production, intelligent dispatching, inspection, and intelligent customer service are all very important application scenarios, she said.
However, after the infrastructure has been provided for the enterprise, the talent is needed to make better use of this infrastructure.
She believes that in the process of the digital upgrade of enterprises, the most lacking is actually a composite talent, which understands both the boundaries of technology and business understanding.
Based on the infrastructure provided by Baidu, has also been committed to building a training system to be able to promote the training of talent in the whole society, said Wu Tian, adding that currently Baidu has trained more than 1 million AI talents, and in the future hope to train more.
Spec performance comparison with Nvidia
from cnTechPost
Baidu AI chip boasts up to 3 times the performance of Nvidia
2020-04-02 20:30:16 GMT+8 | cnTechPost
1
On the afternoon of April 2, Ouyang Jian, General Manager of Baidu Smart Chips, shared the Kunlun chips for the first time in a public class, and disclosed a number of comparison data between the KunlunK200 and Intel T4 GPUs.
Ouyang Jian also showed the killer feature of the Kunlun chip and its compatibility with the China-made processor Feiteng via video.
At the 2018 Baidu AI Developer Conference, Baidu founder, chairman and CEO Robin Li announced the launch of his own AI chip, Kunlun.
Baidu AI chip accumulation was due to its FPGA to do the accumulation of AI acceleration, but also thanks to its software-defined accelerator and XPU architecture years of accumulation.
Baidu first started using FPGAs for AI architecture research and development in 2010, small-scale deployment went online in 2011, more than 10,000 FPGAs were deployed in 2017, AI chips were released in 2018, the second half of 2019, and mass production began in 2020.
Kunlun chips are positioned as general-purpose AI chips, with the goal of providing high-performance, low-cost, highly flexible AI chips.
Ouyang Jian said, "Compared to GPUs, the Kunlun chip has done a good job of being versatile and programmable, and we're still working on making the programmability better."
After the release of Kunlun, news about it was released one after another. Architecture-wise, Kunlun has 2 computing units, 512GB/S of memory bandwidth and 16MB SRAM/unit.
According to Ouyang Jian, 16MB SRAM is good for AI inference, XPU-SDNN on XPU architecture is designed for Tensor and so on, and XPU-Cluster can meet the needs of general processing.
Kunlun's first-generation chips do not use NVLink but are interconnected via the PCIE 4.0 interface. Backed by Samsung's 14nm manufacturing process and 2.5D package, the Kunlun chip can reach peak performance of 260TOPS and power consumption of 150W.
In terms of flexibility and ease of use, Kunlun offers developers a software stack similar to NVIDIA's CUDA, which can be programmed in C/C++ language, making it less difficult for developers to develop.
Currently, based on the first generation of Kunlun chips, Baidu has launched two AI acceleration cards, K100 and K200, the former with twice the arithmetic power and power consumption of the latter.
In today's share, Ouyang Jian gives a series of K200 vs. NVIDIA T4 data, in which the KunlunK200's Benchmark scores over 2,000, more than 3 times that of the NVIDIA T4, under the Gemm-Int8 data type, 4K X 4K matrix.
The Kunlun also has a significant performance advantage under the Bert/Ernie test model commonly used for voice.
In terms of online performance data, the Kunlun's performance is more stable than the NVIDIA T4, and the latency has an advantage.
In the image segmentation YOLOV3 algorithm, Kunlun has an advantage, but the advantage is no longer as obvious. But Ouyang Jian said Baidu is still improving Kunlun's performance through continuous optimization.
He also said that Kunlun has been applied at Baidu's internal scale. As for the external provision of AI computing power, on December 13 last year Baidu provided Kunlun computing power through the Baidu Cloud by way of targeted invitation.
In addition to providing Kunlun's computing power through Baidu Cloud, Ouyang Jian also demonstrated the use of Kunlun acceleration cards in industrial smart devices, demonstrating the use of CPUs and Kunlun acceleration cards for product defect detection, Kunlun can greatly increase speed, but did not give specific comparative data.
Another showcase is Kunlun's killer feature, which is adapted to the China-made processor platform Feiteng.
At the 2019 Feiteng Eco-Partner Conference, Ouyang Jian revealed that KunlunAI chips are adapting to China-made Feiteng servers to do performance tuning work.
In today's online share, Ouyang Jian demonstrates the significant acceleration in image splitting speed that comes with the Kunlun acceleration card.
As the representative of China-made core, Kunlun chose to match with Feteng for the big market of China-made's own chips.
Through the way of Feiteng CPU + KunlunAI accelerator, both parties can better realize the China-made chip in the server market China-made and can also be considered as an important driving force and killer application for the future growth of KunlunAI chip and accelerator card.
from cnTechPost
Baidu AI chip boasts up to 3 times the performance of Nvidia
2020-04-02 20:30:16 GMT+8 | cnTechPost
1
On the afternoon of April 2, Ouyang Jian, General Manager of Baidu Smart Chips, shared the Kunlun chips for the first time in a public class, and disclosed a number of comparison data between the KunlunK200 and Intel T4 GPUs.
Ouyang Jian also showed the killer feature of the Kunlun chip and its compatibility with the China-made processor Feiteng via video.
At the 2018 Baidu AI Developer Conference, Baidu founder, chairman and CEO Robin Li announced the launch of his own AI chip, Kunlun.
Baidu AI chip accumulation was due to its FPGA to do the accumulation of AI acceleration, but also thanks to its software-defined accelerator and XPU architecture years of accumulation.
Baidu first started using FPGAs for AI architecture research and development in 2010, small-scale deployment went online in 2011, more than 10,000 FPGAs were deployed in 2017, AI chips were released in 2018, the second half of 2019, and mass production began in 2020.
Kunlun chips are positioned as general-purpose AI chips, with the goal of providing high-performance, low-cost, highly flexible AI chips.
Ouyang Jian said, "Compared to GPUs, the Kunlun chip has done a good job of being versatile and programmable, and we're still working on making the programmability better."
After the release of Kunlun, news about it was released one after another. Architecture-wise, Kunlun has 2 computing units, 512GB/S of memory bandwidth and 16MB SRAM/unit.
According to Ouyang Jian, 16MB SRAM is good for AI inference, XPU-SDNN on XPU architecture is designed for Tensor and so on, and XPU-Cluster can meet the needs of general processing.
Kunlun's first-generation chips do not use NVLink but are interconnected via the PCIE 4.0 interface. Backed by Samsung's 14nm manufacturing process and 2.5D package, the Kunlun chip can reach peak performance of 260TOPS and power consumption of 150W.
In terms of flexibility and ease of use, Kunlun offers developers a software stack similar to NVIDIA's CUDA, which can be programmed in C/C++ language, making it less difficult for developers to develop.
Currently, based on the first generation of Kunlun chips, Baidu has launched two AI acceleration cards, K100 and K200, the former with twice the arithmetic power and power consumption of the latter.
In today's share, Ouyang Jian gives a series of K200 vs. NVIDIA T4 data, in which the KunlunK200's Benchmark scores over 2,000, more than 3 times that of the NVIDIA T4, under the Gemm-Int8 data type, 4K X 4K matrix.
The Kunlun also has a significant performance advantage under the Bert/Ernie test model commonly used for voice.
In terms of online performance data, the Kunlun's performance is more stable than the NVIDIA T4, and the latency has an advantage.
In the image segmentation YOLOV3 algorithm, Kunlun has an advantage, but the advantage is no longer as obvious. But Ouyang Jian said Baidu is still improving Kunlun's performance through continuous optimization.
He also said that Kunlun has been applied at Baidu's internal scale. As for the external provision of AI computing power, on December 13 last year Baidu provided Kunlun computing power through the Baidu Cloud by way of targeted invitation.
In addition to providing Kunlun's computing power through Baidu Cloud, Ouyang Jian also demonstrated the use of Kunlun acceleration cards in industrial smart devices, demonstrating the use of CPUs and Kunlun acceleration cards for product defect detection, Kunlun can greatly increase speed, but did not give specific comparative data.
Another showcase is Kunlun's killer feature, which is adapted to the China-made processor platform Feiteng.
At the 2019 Feiteng Eco-Partner Conference, Ouyang Jian revealed that KunlunAI chips are adapting to China-made Feiteng servers to do performance tuning work.
In today's online share, Ouyang Jian demonstrates the significant acceleration in image splitting speed that comes with the Kunlun acceleration card.
As the representative of China-made core, Kunlun chose to match with Feteng for the big market of China-made's own chips.
Through the way of Feiteng CPU + KunlunAI accelerator, both parties can better realize the China-made chip in the server market China-made and can also be considered as an important driving force and killer application for the future growth of KunlunAI chip and accelerator card.
from cnTechPost
Huawei Ark Compiler officially supports C language
2020-10-14 21:07:39 GMT+8 | cnTechPost
0
Huawei Ark Compiler has added support for C language, which is the basis for implementing mixed Java and C compilation.
At the end of August 2019, Huawei Ark Compiler was officially open source, taking a leap forward.
For more than a year, the Ark programming system has realized the open-source of the compiler, engine, and debugger, where the compiler's key functions are mainly focused on the static compilation of Java applications.
Huawei stressed that the goal of the Ark project is to build a MapleIR-based cross-language programming environment to achieve cross-language global analysis and optimization.
For example, in the existing Java programming environment, developers often need to write Java and C programs at the same time and call each other through JNI, while the two languages are not directly interoperable.
One of the ultimate visions of the Ark programming system is to provide a consistent representation of different languages through IR, and to optimize globally across languages.
In the future, the Ark engine will not only JIT compile IR code from Java, but will also support JIT compilation of IR code from C and C++.
Code link address.
1,
2.
This open-source content is mainly limited to the scope of O0 capabilities, for C O2 above the optimization of the future for further open source.
Currently, Open Ark Compiler already supports most of the common features of C. The features that are not supported yet are complex types, packed struct, long double, and alloca out-of-scope free.
Currently, it supports aarch64 backend and most of the gcc_torture use cases, such as a third-party collaborative C frontend (open source time to be determined), which can run all C integer use cases of spec2017 correctly.
As a next step, the Ark community plans to do more broadening and deepening work on the Maple compiler.
On the one hand, it will try to support new languages to further validate the language-independence of the Maple base platform, and on the other hand, it will add support for new backends such as RISC-V.
Its community will open-source O2 analysis and optimization phase as soon as possible, performance benchmark LLVM/GCC, while using a unified Maple representation, the layout of cross-language global optimization.
Since the official open-source, Ark community open-source code grew from 50K to 170K, a total of 6.9K Star, and incubate five related subprojects.
Huawei Ark Compiler officially supports C language
2020-10-14 21:07:39 GMT+8 | cnTechPost
0
Huawei Ark Compiler has added support for C language, which is the basis for implementing mixed Java and C compilation.
At the end of August 2019, Huawei Ark Compiler was officially open source, taking a leap forward.
For more than a year, the Ark programming system has realized the open-source of the compiler, engine, and debugger, where the compiler's key functions are mainly focused on the static compilation of Java applications.
Huawei stressed that the goal of the Ark project is to build a MapleIR-based cross-language programming environment to achieve cross-language global analysis and optimization.
For example, in the existing Java programming environment, developers often need to write Java and C programs at the same time and call each other through JNI, while the two languages are not directly interoperable.
One of the ultimate visions of the Ark programming system is to provide a consistent representation of different languages through IR, and to optimize globally across languages.
In the future, the Ark engine will not only JIT compile IR code from Java, but will also support JIT compilation of IR code from C and C++.
Code link address.
1,
2.
This open-source content is mainly limited to the scope of O0 capabilities, for C O2 above the optimization of the future for further open source.
Currently, Open Ark Compiler already supports most of the common features of C. The features that are not supported yet are complex types, packed struct, long double, and alloca out-of-scope free.
Currently, it supports aarch64 backend and most of the gcc_torture use cases, such as a third-party collaborative C frontend (open source time to be determined), which can run all C integer use cases of spec2017 correctly.
As a next step, the Ark community plans to do more broadening and deepening work on the Maple compiler.
On the one hand, it will try to support new languages to further validate the language-independence of the Maple base platform, and on the other hand, it will add support for new backends such as RISC-V.
Its community will open-source O2 analysis and optimization phase as soon as possible, performance benchmark LLVM/GCC, while using a unified Maple representation, the layout of cross-language global optimization.
Since the official open-source, Ark community open-source code grew from 50K to 170K, a total of 6.9K Star, and incubate five related subprojects.
Hendrik_2000
Lieutenant General
The hard part is industrialization of those physics. Indian have alot of scientists and engineers but majority on papers , books, concept, simulation and power point. Dont be like them.
- Status