A recent benchmarking study conducted by Phoronix, a Linux benchmarking website, has revealed the impressive performance gains of Intel’s 5th Generation Xeon Emerald Rapids scalable CPU when AVX-512 instructions are enabled. The tests conducted on a server equipped with two top-end Xeon Platinum 8592+ 64-core CPUs demonstrated an average doubling in performance, with some workloads even achieving performance boosts of over ten times. What’s more remarkable is that these performance gains were achieved without a substantial increase in power consumption.
Phoronix’s benchmarking tests included workloads such as Embree, OpenVKL, and Y-Cruncher, all of which showed notable improvements when AVX-512 was enabled. Particularly, OpenVINO, which supports AVX-VNNI and BF16, showcased the most significant performance gains, making it highly beneficial for AI workloads.
Interestingly, enabling AVX-512 did not significantly impact the peak frequency of the processors. The Xeon Platinum 8592+ CPUs achieved a frequency of 2.95 GHz on all cores when AVX-512 was enabled, compared to 3.01 GHz when it was disabled. Similarly, the 64-core Emerald Rapids chip maintained its 3.9 GHz boost clock regardless of AVX-512 activation.
Although power usage remained relatively unchanged on average, some individual workloads required up to 10% more power when AVX-512 was enabled. The maximum power consumption rose by approximately 120 watts, which is not unexpected considering the tradeoff for increased performance. Additionally, activating AVX-512 resulted in slightly higher temperatures and a minor decrease in frequency, likely due to the higher power draw and increased heat.
Support for AVX-512 instructions is a significant advantage of Intel’s Emerald Rapids CPU. While it may not match the raw performance of AMD’s 4th Generation EPYC Genoa chip with 96 cores, the addition of AVX instructions can substantially alter the landscape, particularly for AI workloads. Microsoft’s choice to pair last-generation Sapphire Rapids chips with AMD’s MI300X GPUs highlights the importance of AVX capabilities in server CPUs.
Overall, the benchmarking results underscore the performance potential of Intel’s 5th Generation Xeon Emerald Rapids CPU when combined with AVX-512 instructions. This breakthrough technology promises substantial performance gains in various workloads, making it an attractive choice for AI and other compute-intensive applications.
The source of the article is from the blog tvbzorg.com