Linux's Cache Aware Scheduling On AMD Ryzen 9 9950X3D 3D V-Cache

Linux’s Cache Aware Scheduling On AMD Ryzen 9 9950X3D 3D V-Cache

Last updated: 2025/12/29 at 7:42 AM

News Room Published 29 December 2025

One of the many interesting Linux kernel innovations I have closely been following this year has been the proposed Cache Aware Scheduling support. I have shown the Cache Aware Scheduling performance on AMD EPYC as well as the Intel Xeon 6 Granite Rapids performance, but what about desktops? In this article is a quick look at Cache Aware Scheduling with the Ryzen 9 9950X3D.

While not as interesting as running on big server CPUs, stemming from my own curiosity and reader interest over the holiday weekend I ran benchmarks looking at the performance of Linux 6.18 upstream compared to applying the Cache Aware Scheduling v2 patches on top.

AMD Ryzen 9 9950X3D box

The AMD Ryzen 9 9950X3D was the focus of this round of benchmarking.

No other software/hardware changes were made besides changing out the kernel in use with the carried Cache Aware Scheduling “CAS” patches.

Selenium benchmark with settings of Benchmark: Speedometer 3.1, Browser: Firefox. CAS v2 was the fastest.

Selenium benchmark with settings of Benchmark: Jetstream 2, Browser: Firefox. Mainline 6.18 was the fastest.

Darktable benchmark with settings of Test: Server Room, Acceleration: CPU-only. Mainline 6.18 was the fastest.

RawTherapee benchmark with settings of Total Benchmark Time. Mainline 6.18 was the fastest.

SVT-AV1 benchmark with settings of Encoder Mode: Preset 8, Input: Bosphorus 4K. Mainline 6.18 was the fastest.

SVT-JPEG-XS benchmark with settings of Input: Bosphorus 4K. Mainline 6.18 was the fastest.

For various desktop-type workloads, the Cache Aware Scheduling kernel didn’t yield any gains for the AMD Ryzen 9 9950X3D Linux system.

Timed Linux Kernel Compilation benchmark with settings of Build: defconfig. CAS v2 was the fastest.

Timed Linux Kernel Compilation benchmark with settings of Build: allmodconfig. Mainline 6.18 was the fastest.

DaCapo Benchmark benchmark with settings of Java Test: Avrora AVR Simulation Framework. CAS v2 was the fastest.

DaCapo Benchmark benchmark with settings of Java Test: Apache Tomcat. CAS v2 was the fastest.

DaCapo Benchmark benchmark with settings of Java Test: Tradesoap. CAS v2 was the fastest.

DaCapo Benchmark benchmark with settings of Java Test: Zxing 1D/2D Barcode Image Processing. CAS v2 was the fastest.

Most workloads tested were rather unchanged for this dual CCD AMD Ryzen 9 9950X3D processor compared to testing on the bigger server CPUs.

Apache Cassandra benchmark with settings of Test: Writes. CAS v2 was the fastest.

Pogocache benchmark with settings of Set To Get Ratio: 1:5. CAS v2 was the fastest.

nginx benchmark with settings of Connections: 100. CAS v2 was the fastest.

ClickHouse benchmark with settings of 100M Rows Hits Dataset, First Run / Cold Cache. Mainline 6.18 was the fastest.

Among Linux server workloads on the Ryzen 9 9950X3D — such as if using this CPU in a development box, SOHO server, or akin to the EPYC 4000 series — most workloads also didn’t see any real difference on the patched kernel.

PostgreSQL benchmark with settings of Scaling Factor: 100, Clients: 50, Mode: Read Write. CAS v2 was the fastest.

PostgreSQL benchmark with settings of Scaling Factor: 100, Clients: 50, Mode: Read Write, Average Latency. CAS v2 was the fastest.

PostgreSQL benchmark with settings of Scaling Factor: 100, Clients: 100, Mode: Read Write. CAS v2 was the fastest.

PostgreSQL benchmark with settings of Scaling Factor: 100, Clients: 100, Mode: Read Write, Average Latency. CAS v2 was the fastest.

PostgreSQL was one of the notable exceptions though in seeing improved performance — better throughput and lower latency — on the Cache Aware Scheduling enabled kernel for this AMD Ryzen 9 system.

Ethr benchmark with settings of Server Address: localhost, Protocol: TCP, Test: Bandwidth, Threads: 4. CAS v2 was the fastest.

Ethr benchmark with settings of Server Address: localhost, Protocol: TCP, Test: Latency, Threads: 8. CAS v2 was the fastest.

Some of the localhost network testing did benefit too from the Cache Aware Scheduling on this AMD Zen 5 3D V-Cache desktop CPU.

Xonotic benchmark with settings of Resolution: 3840 x 2160, Effects Quality: Ultimate. Mainline 6.18 was the fastest.

Tesseract benchmark with settings of Resolution: 3840 x 2160. CAS v2 was the fastest.

GravityMark benchmark with settings of Resolution: 1920 x 1080, Renderer: Vulkan. CAS v2 was the fastest.

GPUScore: Breaking Limit benchmark with settings of Resolution: 3840 x 2160, Ray-Tracing: Off. Mainline 6.18 was the fastest.

ParaView benchmark with settings of Test: Wavelet Volume, Frames: 3000, Resolution: 1920 x 1080. CAS v2 was the fastest.

Overall the Cache Aware Scheduling kernel tests on the Ryzen 9 9950X3D weren’t as exciting as when running on big AMD EPYC or Intel Xeon servers. Now at least you know.