Intel Xeon Max Sees Some Performance Gains For OpenVINO & ONNX With Linux 6.9

Written by Michael Larabel in Software on 4 April 2024 at 10:30 AM EDT. Page 3 of 5. 2 Comments.
Timed FFmpeg Compilation benchmark with settings of Time To Compile. Linux 6.8 was the fastest.
Timed Linux Kernel Compilation benchmark with settings of Build: defconfig. Linux 6.9-rc2 was the fastest.
Timed Linux Kernel Compilation benchmark with settings of Build: allmodconfig. Linux 6.8 was the fastest.
oneDNN benchmark with settings of Harness: IP Shapes 3D, Engine: CPU. Linux 6.8 was the fastest.
oneDNN benchmark with settings of Harness: Convolution Batch Shapes Auto, Engine: CPU. Linux 6.9-rc2 was the fastest.
oneDNN benchmark with settings of Harness: Deconvolution Batch shapes_1d, Engine: CPU. Linux 6.9-rc2 was the fastest.
oneDNN benchmark with settings of Harness: Deconvolution Batch shapes_3d, Engine: CPU. Linux 6.9-rc2 was the fastest.
oneDNN benchmark with settings of Harness: Recurrent Neural Network Training, Engine: CPU. Linux 6.8 was the fastest.

With the recurrent neural network training test of the Intel oneDNN library there was a big regression when running on 6.9-rc2.

TensorFlow benchmark with settings of Device: CPU, Batch Size: 32, Model: AlexNet. Linux 6.8 was the fastest.
TensorFlow benchmark with settings of Device: CPU, Batch Size: 64, Model: AlexNet. Linux 6.8 was the fastest.
TensorFlow benchmark with settings of Device: CPU, Batch Size: 1, Model: GoogLeNet. Linux 6.8 was the fastest.
TensorFlow benchmark with settings of Device: CPU, Batch Size: 1, Model: ResNet-50. Linux 6.9-rc2 was the fastest.
TensorFlow benchmark with settings of Device: CPU, Batch Size: 32, Model: ResNet-50. Linux 6.8 was the fastest.
TensorFlow benchmark with settings of Device: CPU, Batch Size: 256, Model: ResNet-50. Linux 6.9-rc2 was the fastest.

TensorFlow wasn't seeing much of a difference on the new kernel.

Stress-NG benchmark with settings of Test: Hash. Linux 6.9-rc2 was the fastest.
Stress-NG benchmark with settings of Test: MMAP. Linux 6.9-rc2 was the fastest.
Stress-NG benchmark with settings of Test: NUMA. Linux 6.9-rc2 was the fastest.
Stress-NG benchmark with settings of Test: Poll. Linux 6.8 was the fastest.
Stress-NG benchmark with settings of Test: Mutex. Linux 6.9-rc2 was the fastest.
Stress-NG benchmark with settings of Test: Pthread. Linux 6.8 was the fastest.
Stress-NG benchmark with settings of Test: IO_uring. Linux 6.8 was the fastest.
Stress-NG benchmark with settings of Test: SENDFILE. Linux 6.9-rc2 was the fastest.
Stress-NG benchmark with settings of Test: Floating Point. Linux 6.8 was the fastest.
Stress-NG benchmark with settings of Test: Mixed Scheduler. Linux 6.8 was the fastest.
Stress-NG benchmark with settings of Test: Socket Activity. Linux 6.8 was the fastest.

Unlike the AMD EPYC Genoa-X results where some of the Stress-NG micro-benchmarks were pulling some nice wins on Linux 6.9, the Xeon Max performance was rather flat.


Related Articles
Intel Core Ultra "Meteor Lake" Performance Improves With Linux 6.9
GCC 14 vs. LLVM Clang 18 Compiler Performance On Fedora 40
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
Linux 6.9 Features: DM VDO, AMD Preferred Core, Intel FRED & Larger Console Fonts