64K Kernel Page Size Performance Benefits For HPC Shown With NVIDIA's GH200 Grace CPU

Written by Michael Larabel in Software on 27 February 2024 at 12:00 PM EST. Page 2 of 5. 9 Comments.
Rodinia benchmark with settings of Test: OpenMP LavaMD. EPYC 9754 2P was the fastest.

The Linux 6.5 to 6.8 upgrade on the GH200 didn't tend to make too much of a difference while quite quickly the 64K kernel build began to show better performance over the 4K kernel.

Algebraic Multi-Grid Benchmark benchmark with settings of . EPYC 9684X 2P was the fastest.

In some workloads the 64K kernel increased the CPU performance by several percent but not necessarily enough to change the outcome in regards to the EPYC and Xeon CPU positioning.

NWChem benchmark with settings of Input: C240 Buckyball. EPYC 9554 2P was the fastest.
Xcompact3d Incompact3d benchmark with settings of Input: X3D-benchmarking input.i3d. Xeon Platinum 8592+ 2P was the fastest.

Seeing several percent improvements isn't out of the ordinary for the ARM64 4K vs. 64K performance and jives with testing I've done on other ARM server hardware in the past like with Ampere Computing.

LULESH benchmark with settings of . Xeon Platinum 8592+ 2P was the fastest.

The LULESH hydrodynamics software saw a huge boost in performance out of the 64K page size kernel build.


Related Articles