After AMD Connect introduced the 32-core and 64-core EPYC Dragon processors, CPU performance began to move toward stake. Intel’s native-architecture processors are incomparable at this point, with mainstream desktops having the most eight cores and UE processors having up to 28 cores, but Intel believes cpu cores are not the only metric, but rather serving customers from a workload perspective.
Lisa A. Spelman, Vice President of Intel’s Data Center Division and General Manager of Processor stoprocessor and data center marketing, spoke to the media about Intel’s business priorities in data centers, 5G, AI, cloud services, and more.
Lisa A. Spelman says that in 2020 Intel will continue to promote second-generation Xeon Scalable Processors, and will launch a third-generation Xeon Extensible Processor, also known as Cooper Lake’s 14nm Ethel, that will bring a new DL Boost instruction set to accelerate AI performance. Mainly improve the training performance.
Lisa A. Spelman says Intel doesn’t just look at how much core and how high frequencies a chip has, but looks at the whole, serving customers from the perspective of actual load and customer needs, having previously Intel spent more than a year communicating with customers on the second-generation Saee Scalable Processor to improve the user experience.
Lisa A. Spelman points out that customers can compare the reality, one is Intel’s product, one is a friend’s product, the core number is more and the frequency is the same, they are used in the same AI or other application scenarios, you can see the built-in AI acceleration of the Xeon processor will perform better.
In addition, Lisa A. Spelman mentions that Intel works with customers to optimize the AVX512 instruction set, especially in the high-performance computing market, to ensure that the customer’s software can take advantage of the AVX512 instruction set and ultimately see significant performance differences.