Intel bets on a secret weapon to beat AMD in some AI workloads — AVX-512 instructions will help Xeon beat Epyc in specific scenarios but Intel will have to work much, much harder to worry AMD

Key Takeaways:

– Intel’s 5th-Gen Xeon Scalable processors, built on the Emerald Rapids architecture, can benefit from a significant performance boost when running AVX-512 workloads.
– AVX-512 is an instruction set that improves CPU performance by running the same instructions across different data points.
– Recent generations of Intel Xeon CPUs have addressed concerns about poor power efficiency and excessive heat generation associated with AVX-512.
– Benchmarking site Phoronix found that enabling AVX-512 almost doubled the average performance of an Intel Xeon Platinum 8592+ CPU in specific workloads.
– AVX-512 particularly saw a significant performance increase in the OpenVINO benchmark, which is an Intel toolkit for neural networks.
– The CPU did not experience major power usage or heat generation issues when using AVX-512.
– However, Intel’s 5th-gen Xeon CPU does not match the performance of AMD’s Ryzen Threadripper Pro 7995WX, which holds the world record in several categories.

TechRadar:

Intel’s 5th-Gen Xeon Scalable processors, built on the Emerald Rapids architecture, can benefit from a huge raw performance boost when running AVX-512 workloads.

An Intel Xeon Platinum 8592+ CPU can almost double its average performance while also increasing power efficiency when running AVX-512 instructions in a handful of specific workloads, according to benchmarking site Phoronix

Source link

AI Eclipse TLDR:

Intel’s 5th-Gen Xeon Scalable processors, based on the Emerald Rapids architecture, can experience a significant performance boost when running AVX-512 workloads. According to benchmarking site Phoronix, an Intel Xeon Platinum 8592+ CPU can nearly double its average performance while also improving power efficiency when running AVX-512 instructions in specific workloads. AVX-512 is an instruction set that enhances CPU performance by executing the same instructions across different data points. It aims to accelerate tasks like data compression, cryptographic calculations, and image processing. Although AVX-512 was previously criticized for poor power efficiency and heat generation, recent generations of Intel Xeon CPUs have addressed these concerns. Phoronix conducted tests on two 64-core Intel Xeon Platinum 8592+ CPUs and found that enabling AVX-512 resulted in a doubled performance in various benchmarks, including Embree, OpenVKL, and Y-Cruncher. In the OpenVINO benchmark, AVX-512 increased performance by over three times. OpenVINO is an Intel toolkit designed for neural networks and enables developers to create AI-centric applications for tasks like vision emulation, speech recognition, and natural language processing. The CPU experienced no significant power usage or heat generation issues, highlighting the benefits of AVX-512 in today’s server-oriented processors. However, while Intel’s latest 5th-gen Xeon CPU is considered one of the best for servers, it does not match the performance of AMD’s 96-core Ryzen Threadripper Pro 7995WX, which currently holds world records in multiple categories.