Benchmarking The Performance Impact To AMD Inception Mitigations

Written by Michael Larabel in Software on 15 August 2023 at 10:30 AM EDT. Page 2 of 8. 35 Comments.
NAMD benchmark with settings of ATPase Simulation, 327,506 Atoms. safe RET was the fastest.
Algebraic Multi-Grid Benchmark benchmark with settings of . off was the fastest.
Timed MrBayes Analysis benchmark with settings of Primate Phylogeny Analysis. off was the fastest.
OpenFOAM benchmark with settings of Input: drivaerFastback, Medium Mesh Size, Execution Time. off was the fastest.

The good news is that the HPC performance on EPYC servers was largely unchanged by the new AMD Inception mitigations on this EPYC server.

OpenRadioss benchmark with settings of Model: Cell Phone Drop Test. off was the fastest.
OpenRadioss benchmark with settings of Model: Rubber O-Ring Seal Installation. off was the fastest.
OpenRadioss benchmark with settings of Model: Bird Strike on Windshield. off was the fastest.

The OpenRadioss software based on Altair Radioss did see some performance regressions primarily when using the alternative IBPB mitigation mode... But again, "safe RET" is the default, which for some OpenRadioss models did show some minor setbacks in performance.

SPECFEM3D benchmark with settings of Model: Mount St. Helens. off was the fastest.
SPECFEM3D benchmark with settings of Model: Layered Halfspace. safe RET was the fastest.
SPECFEM3D benchmark with settings of Model: Homogeneous Halfspace. off was the fastest.

On the whole though for the high performance computing workloads the AMD Inception mitigations didn't change much there and at least not erode the hefty margins AMD has generally had over the Intel Xeon competition.

Embree benchmark with settings of Binary: Pathtracer ISPC, Model: Crown. off was the fastest.
Embree benchmark with settings of Binary: Pathtracer ISPC, Model: Asian Dragon. safe RET was the fastest.
OpenVKL benchmark with settings of Benchmark: vklBenchmark ISPC. off was the fastest.
OSPRay benchmark with settings of Benchmark: particle_volume/scivis/real_time. off was the fastest.

Rendering workloads like Embree also were not impacted.


Related Articles
Intel Core Ultra "Meteor Lake" Performance Improves With Linux 6.9
GCC 14 vs. LLVM Clang 18 Compiler Performance On Fedora 40
Intel Xeon Max Sees Some Performance Gains For OpenVINO & ONNX With Linux 6.9
Linux 6.9 Drives AMD 4th Gen EPYC Performance Even Higher For Some Workloads
Intel Core Ultra "Meteor Lake" Yields Faster Performance With Linux 6.9
Google Cloud's C3D Instances Provide Strong Performance Value For PingCAP's TiDB