Early Linux 6.17 Tests Show Some AMD Strix Halo Performance Improvements & Regressions

Written by Michael Larabel in Linux Kernel on 11 August 2025 at 03:55 PM EDT. 5 Comments
LINUX KERNEL
Even prior to the Linux 6.17-rc1 release on Sunday I already had kicked off some Linux 6.17 Git benchmarking in being eager to see how the performance is beginning to shape up for this next kernel release that is set to power the likes of Ubuntu 25.10 and Fedora 43. There is some good news and bad news with my early testing on the ZBook Ultra G1a for AMD Strix Halo.

This weekend I started running some initial benchmarks to see how Linux 6.17 Git is looking relative to Linux 6.16. Starting off with the AMD Ryzen AI Max+ PRO 395 I am seeing some workloads having some nice improvements on the new kernel but also some regressions. Now that Linux 6.17-rc1 is out, I will be testing on more benchmarks and more hardware over the coming days and weeks.
Linux 6.17 Git Smoke Test Strix Halo

Some of the initial takeaways of this first Linux 6.16 vs. 6.17 Git testing include:
Selenium benchmark with settings of Benchmark: PSPDFKit WASM, Browser: Firefox. Linux 6.17 Git 8 Aug was the fastest.

Schbench benchmark with settings of Message Threads: 32, Cache Footprint: 256 kb, Locking: Yes. Linux 6.17 Git 8 Aug was the fastest.

In a few web browser tests and also schbench as the Linux kernel synthetic scheduler benchmark was some really nice improvements. Particularly due to the strong difference in the single-threaded web browser performance, coming to mind is the AMD Hardware Feedback Interface that was merged for Linux 6.17. I haven't yet verified if the AMD Hardware Feedback Interface is responsible for these improvements or regressions but it's something that comes to mind on the AMD side with Linux 6.17.
Unvanquished benchmark with settings of Resolution: 2880 x 1800, Effects Quality: Medium. Linux 6.16 was the fastest.

Unvanquished benchmark with settings of Resolution: 2880 x 1800, Effects Quality: High. Linux 6.16 was the fastest.

ParaView benchmark with settings of Test: Wavelet Volume, Frames: 3000, Resolution: 1920 x 1080. Linux 6.16 was the fastest.

ParaView benchmark with settings of Test: Wavelet Volume, Frames: 600, Resolution: 1920 x 1080. Linux 6.16 was the fastest.

In some of the graphics tests there were regressions compared to Linux 6.17, which could just be a separate AMDGPU regression as well. Again, testing on more hardware over the coming days at Phoronix.
yquake2 benchmark with settings of Renderer: Vulkan, AF: Off, MSAA: Off, Resolution: 2560 x 1440. Linux 6.17 Git 8 Aug was the fastest.

Renaissance benchmark with settings of Test: In-Memory Database Shootout. Linux 6.16 was the fastest.

yquake2 benchmark with settings of Renderer: Vulkan, AF: On, MSAA: Off, Resolution: 2560 x 1440. Linux 6.17 Git 8 Aug was the fastest.

In a few lightweight graphics tests was better performance on Linux 6.17.
Stress-NG benchmark with settings of Test: System V Message Passing. Linux 6.16 was the fastest.

Stress-NG benchmark with settings of Test: Forking. Linux 6.16 was the fastest.

Some hits in the Stress-NG kernel micro-benchmarks.
Flexible IO Tester benchmark with settings of Type: Random Write, Engine: POSIX AIO, Direct: Yes, Block Size: 4KB, Job Count: 32, Disk Target: Default Test Directory. Linux 6.17 Git 8 Aug was the fastest.

Flexible IO Tester benchmark with settings of Type: Random Write, Engine: POSIX AIO, Direct: Yes, Block Size: 4KB, Job Count: 16, Disk Target: Default Test Directory. Linux 6.17 Git 8 Aug was the fastest.

Flexible IO Tester benchmark with settings of Type: Random Read, Engine: IO_uring, Direct: No, Block Size: 4KB, Job Count: 32, Disk Target: Default Test Directory. Linux 6.17 Git 8 Aug was the fastest.

Flexible IO Tester benchmark with settings of Type: Random Read, Engine: IO_uring, Direct: No, Block Size: 4KB, Job Count: 1, Disk Target: Default Test Directory. Linux 6.17 Git 8 Aug was the fastest.

Apache Cassandra benchmark with settings of Test: Writes. Linux 6.17 Git 8 Aug was the fastest.

RocksDB benchmark with settings of Test: Read While Writing. Linux 6.17 Git 8 Aug was the fastest.

But also some nice I/O improvements observed with the early Linux 6.17 kernel state.
Stress-NG benchmark with settings of Test: IO_uring. Linux 6.17 Git 8 Aug was the fastest.

IO_uring is looking even better in Linux 6.17.
nginx benchmark with settings of Connections: 200. Linux 6.16 was the fastest.

Stay tuned for more Linux 6.17 kernel benchmarking on different hardware in seeing if some of the improvements and regressions are just limited to Strix Halo or isolated enhancements or more broad... At least from the I/O side the Linux 6.17 results were most promising as well as some of the other improvements but also the regressions may be concerning if widespread.
Related News
About The Author

Michael Larabel is the principal author of Phoronix.com and founded the site in 2004 with a focus on enriching the Linux hardware experience. Michael has written more than 20,000 articles covering the state of Linux hardware support, Linux performance, graphics drivers, and other topics. Michael is also the lead developer of the Phoronix Test Suite, Phoromatic, and OpenBenchmarking.org automated benchmarking software. He can be followed via Twitter, LinkedIn, or contacted via MichaelLarabel.com.

Popular News This Week