AMD EPYC 7343 / EPYC 7443 Linux Performance

When it came to looking at the performance across the code compilation speed, the dual EPYC 7343 configuration and above could outperform a single Xeon Platinum 8380. Not too surprising as the AVX-512 and other instruction advantages of Ice Lake don't come into play here but principally the core count. The code compilation performance was looked at while compiling the Linux kernel, ImageMagick, LLVM, FFmpeg, Godot, WASMer, Node.js, and Mesa.

Meanwhile for various "creator" workloads, the Xeon Platinum 8380 does come out ahead of the EPYC 7343 2P configuration. Among the creator workloads benchmarked were Blender, POV-Ray, Appleseed, LuxCore, V-RAY, SVT-HEVC, SVT-AV1, Embree, oneDNN, OpenVKL, ASTCENC, BRL-CAD, and others.

Here's a look at the overall HPC performance relative to the different processors being benchmarked. Among the high performance computing tests used were NAS Parallel Benchmarks, Rodinia, AMG, NAMD, GROMACS, NWChem, LAMMPS, Pennant, OpenFOAM, PlaidML, and WRF.

When getting into the machine learning workloads is where the Intel Xeon Scalable Ice Lake processors offer much greater competition thanks to AVX-512 / DL-BOOST.

Or when looking at workloads leveraging Intel oneAPI components where they are well tuned for Intel Ice Lake capabilities, the Xeon Platinum 8380 with just 40c/80t is able to take on the high-end Milan processors. The oneAPI tests here are Embree, oneDNN, Open Image Denoise, and OpenVKL.

The AMD EPYC 7003 series going up to 64 cores compared to 40 cores with Ice Lake pays off especially for the rendering workloads. Rendering benchmarks used for this article were V-RAY, LuxCore, Appleseed, Blender, rays1bench, POV-Ray, Tachyon, and C-Ray.