For gamers using Linux on older hardware, a new kernel patch series could be a game-changer. Linux kernel developer Peter Zijlstra has posted the second version of a patch set titled "sched: Flatten the pick," which aims to fix critical inefficiencies in how the Linux scheduler handles cgroup scheduling. This fix could breathe new life into aging systems, but it also has implications for modern high-core-count machines. Let's dive into the details.
Understanding the Scheduler Problem
The issue revolves around the way the Linux scheduler distributes CPU time among task groups (cgroups). In a multi-core system, a group's total weight—a measure of its priority—gets fragmented across every available CPU core. Zijlstra described this problem as "a pain in the ar*e" and traced it to a formula that splits a task group's overall weight proportionally across all CPUs.
Weight Fragmentation Across CPUs
On a 64-core machine, the weight assigned to a cgroup per CPU effectively becomes that of a process with a nice value of 19 (very low priority). On a 256-core system, which is common in servers today, the per-CPU weight becomes even tinier. This fragmentation means that a group's tasks—such as a game and its background processes—can lose priority badly when competing against other workloads, leading to stuttering and unplayable performance.
The Two-Part Issue
Zijlstra breaks the problem into two components. The first part is that the usual fix—inflating the group weight by the number of CPUs—backfires when all of a group's load lands on a single CPU. The weight then balloons well beyond the range of nice -20 (the highest priority), causing the math to break down. The second part concerns how the scheduler selects the next task to run: it currently must step through multiple cgroup levels to make its choice. The proposed patch collapses these levels into a single step, simplifying and speeding up the decision process.
Experimental Benchmark: Old Hardware, Dramatic Improvement
To demonstrate the impact, Zijlstra ran a stress test using an older configuration: an Intel Core i7-2600K CPU and an AMD Radeon RX 580 graphics card. He loaded up Shadows: Awakening from GOG via Lutris, using GE-Proton10-34 and Steam Runtime 3 (sniper). To simulate real-world competition for CPU time, he launched eight "spinner" processes (CPU-intensive loops) alongside the game—one per CPU thread. Initially, the game went from playable to "almost unplayable, as in proper terrible."
He then restarted the test with a shorter scheduler time slice, set to one-tenth of the default using the chrt command, and recorded both sessions with MangoHud. The results were striking:
- FPS min: Default slice 3.8 → Shorter slice 20.6
- FPS avg: Default slice 48.0 → Shorter slice 57.2
- FPS max: Default slice 87.4 → Shorter slice 80.3
- Frame time min (ms): Default 9.4 → Shorter 8.4
- Frame time avg (ms): Default 34.5 → Shorter 19.5
- Frame time max (ms): Default 107.4 → Shorter 37.2
The shorter slice dramatically improved minimum FPS and reduced frame time spikes, making the game smooth again. Zijlstra noted that he hasn't compared against a kernel without the "flat" scheduling yet—this was just a sanity check to ensure the fix works under non-trivial workloads.
What This Means for Linux Gamers
Real-World Stress Similar to Normal Use
The stress test may seem extreme, but it closely mirrors typical gaming scenarios. When you game, background tasks like Discord, a web browser with a dozen tabs, or a system update all compete for CPU time—much like those spinner processes. The scheduler's weight fragmentation issue can turn a playable game into a slideshow on older hardware, exactly as Zijlstra observed.
Benefits for Modern High-Core Systems
While the test used an aging i7-2600K (4 cores, 8 threads), the patch's benefits are not limited to old PCs. The weight fragmentation problem scales with core count: more cores make the per-CPU weight smaller and the priority crippling worse. A system with 16 or more modern cores could see just as much gain, if not more, once the patch lands. Gaming workloads often involve multiple threads that compete with background services, and this fix could lead to smoother frame rates across the board.
Not Yet in Mainline Kernel
As of now, the patch series has not been merged into the mainline Linux kernel. It still requires review from the relevant kernel maintainers and will likely undergo several revisions before it appears in a stable release. However, the clear performance data and the fundamental nature of the fix make it a promising candidate for inclusion in a future kernel version.
If you game on Linux—especially on older or high-core-count hardware—this development is worth watching. Keep an eye on kernel release notes and the linux-kernel mailing list for updates on "sched: Flatten the pick." In the meantime, you can explore alternative ways to improve gaming scheduler performance on your system, such as adjusting CPU governors or using real-time priorities. For more details, visit the original discussion on the kernel mailing list or the Phoronix coverage linked in the source article.