AVX-512 Performance + Power Efficiency Shines With AMD Strix Halo

Written by Michael Larabel in Processors on 28 May 2025 at 12:30 PM EDT. Page 5 of 8. 22 Comments.
ONNX Runtime benchmark with settings of Model: fcn-resnet101-11, Device: CPU, Executor: Standard. AVX-512 On was the fastest.
ONNX Runtime benchmark with settings of Model: super-resolution-10, Device: CPU, Executor: Standard. AVX-512 On was the fastest.
ONNX Runtime benchmark with settings of Model: bertsquad-12, Device: CPU, Executor: Standard. AVX-512 On was the fastest.
ONNX Runtime benchmark with settings of Model: ArcFace ResNet-100, Device: CPU, Executor: Standard. AVX-512 On was the fastest.
ONNX Runtime benchmark with settings of Model: ResNet50 v1-12-int8, Device: CPU, Executor: Standard. AVX-512 On was the fastest.
ONNX Runtime benchmark with settings of Model: ResNet101_DUC_HDC-12, Device: CPU, Executor: Standard. AVX-512 On was the fastest.

For the last of the AVX-512 AI workloads, a very nice showing for Strix Halo AVX-512 performance with Microsoft's ONNX Runtime on Ubuntu Linux with this AMD Ryzen AI MAX+ PRO 395 laptop.

OSPRay benchmark with settings of Benchmark: gravity_spheres_volume/dim_512/ao/real_time. AVX-512 On was the fastest.
OSPRay benchmark with settings of Benchmark: gravity_spheres_volume/dim_512/scivis/real_time. AVX-512 On was the fastest.
OSPRay benchmark with settings of Benchmark: gravity_spheres_volume/dim_512/pathtracer/real_time. AVX-512 On was the fastest.

Beyond the AI workloads, AVX-512 with Strix Halo and Zen 5 at large is very beneficial for other workloads too like the OSPRay renderer.

OSPRay Studio benchmark with settings of Camera: 1, Resolution: 1080p, Samples Per Pixel: 1, Renderer: Path Tracer, Acceleration: CPU. AVX-512 On was the fastest.
OSPRay Studio benchmark with settings of Camera: 1, Resolution: 1080p, Samples Per Pixel: 16, Renderer: Path Tracer, Acceleration: CPU. AVX-512 On was the fastest.
OSPRay Studio benchmark with settings of Camera: 1, Resolution: 1080p, Samples Per Pixel: 32, Renderer: Path Tracer, Acceleration: CPU. AVX-512 On was the fastest.
OSPRay Studio benchmark with settings of Camera: 1, Resolution: 1080p, Samples Per Pixel: 32, Renderer: Path Tracer, Acceleration: CPU. AVX-512 On was the fastest.
OSPRay Studio benchmark with settings of Camera: 1, Resolution: 1080p, Samples Per Pixel: 32, Renderer: Path Tracer, Acceleration: CPU. AVX-512 On was the fastest.

Very significant uplift without any increased SoC power consumption.

Related Articles
NVIDIA Vera SMT Performance Benchmarks
Intel Granite Rapids WS vs. AMD Threadripper Zen 5 SMT Performance
Intel Xeon 600 Workstation Performance vs. AMD Threadripper 9000 In Nearly 400 Benchmarks
Intel Granite Rapids WS Core Scaling Performance With The Xeon 678X
Intel Hyper Threading Performance On The Xeon 678X "Granite Rapids-WS"
Intel AMX Performance With The Xeon 678X Workstation CPU