AMD EPYC 9004 ‘Genoa’ CPUs Deliver a 35% Performance Boost With AVX-512 Enabled & At The Same Power
The AMD EPYC 9654 processor is one of a series of new server processors that have been hailed “the fastest server CPUs on the planet” upon launch, and Michael Larabel of Phoronix has put the new fourth-gen Genoa CPUs to the test in an impressive 130 benchmarks in the Ubuntu 22.10 OS environment. Nothing was left unturned in these benchmarks — performance, temperature, frequency limits, and more, were tested to find out how the new EPYC processor handles the recent implementation of AVX-512 that was added to this new processor series. AMD Zen 4 introduced the AVX-512 instruction set, first presented by Intel and incorporated into the company’s Intel Xeon Phi x200, Skylake-X, and the recent Xeon Scalable processors. Each extension in the AVX-512 instruction set must be independently executed. Recently, AVX-512 has seen use in several instances, such as increasing performance. With AMD, incorporating AVX-512 is stated to contain enhanced performance and managing data in processing video, analyzing financial equations, and simulations in scientific advancements. Larabel has tested other AMD processors with AVX-512 active, such as the Ryzen 9 7950X and the EPYC 9004 series. In his previous tests, AVX-512 was quite favorable to both processors, showing increased efficiency while keeping consumption and clock frequencies lower, especially in large workloads. He used the AMD EPYC 9654 2P processor for his recent test with AVX-512 active and deactivated in Ubuntu 22.10, which uses the current Linux kernel (v6.1). In his benchmarks surrounding artificial intelligence, the performance with AVX-512 active performed thirty-five percent higher (if not more in some cases) than with the instruction set deactivated. Processor power consumption in AI workloads was very close to negligible, but while active, the AVX-512 instances were still better by keeping power consumption levels lower. One AI-related set of benchmarks, the Neural Magic DeepSparse 1.1, did show welcoming results for AVX-512 in the new AMD EPYC 9654 processor, but it was not as dramatic as some other machine learning workload tests. Neural Magic DeepSparse is “a sparsity-aware inference runtime” that offers graphics processing performance on processors and APIs, allowing the integration of machine learning. You can find out more about it here. Another AI-based benchmark, Mobile Neural Network 2.1, was an “odd duck” for the flurry of benchmarks as the AVX-512 implementation performed worse, and it was only in one specific test with the model “Inception-v3.” Larabel mentions that it is possible that the software itself could be the catalyst but does not have a definite answer. Crypto benchmarks and Tencent’s NCNN models were favorable, so the author switched to Intel-specific software focused on the benefits of AVX-512. Again, AMD EPYC outshined in AVX-512-enabled tests. While running the Intel Open Image Denoise (v1.4.0) benchmark, two instances showed negligible results, but Larabel showed that the power consumption was still lower with the active AVX-512. Larabel ends his testing for now but notes that AMD’s Zen 4 architecture continues to show favorable results for this new generation compared to the current Intel Xeon Scalable processors and it looks like even the upcoming Sapphire Rapids Xeon chips will have a hard time competing against Genoa CPUs. News Source: Phoronix
