Earlier this week, Nvidia announced the first Arm-based Grace processor for artificial intelligence and HPC servers. Although this move was expected by many market observers. The debut of AMD had a negative impact on the stock prices of AMD and Intel. Nvidia is undoubtedly a strong competitor, but Intel believes that it does not really pose an imminent threat to its data center business.
Intel’s official said,” In Ice Lake In China, we have extraordinary expansion in AI capabilities. This is not our response to them. Obviously, this kind of artificial intelligence-enhanced CPU idea is an area where Intel is far ahead. Obviously, what I want to say is that the concept of CPU is proposed by Intel. We are now integrating AI into it, and we expect this to be our future offensive rather than defensive area. ”
Nvidia claims that the Grace processor for AI and HPC machines will be more than 10 times faster in AI and selected HPC workloads compared to current x86 processors, but it will not be available until early 2023. Such performance expectations are quite scary, but competitors AMD and Intel are not standing still.
Moreover, Intel’s latest Xeon processor integrates AI acceleration technology under the general name DL Boost, and currently includes instruction set extensions such as AVX512_VNNI and AVX512_BF16. Facts have proved that these two products are quite competitive overall. When dealing with optimized algorithms, they can make Intel’s CPU performance surpass Nvidia’s GPU by as much as 15 times.
In general, Arm, AMD, Intel, Nvidia, and a series of other companies are working hard to improve the performance of their processors include CPU, GPU, IPU, VPU, etc. in AI and HPC workloads because of the impact of artificial intelligence. The demand for supercomputing is rapidly increasing. Therefore, the progress in artificial intelligence and HPC in the next few years will exceed the previous 30 years.
|VIA|
