Linux 7.3 Scheduler Changes Target Old Gaming Hardware, Hybrid CPUs and Short-Task Latency

Updated Oct. 26, 2026, 3:29 a.m. ET: Adds that Linux 7.3 shipped as a stable kernel on Oct. 18.

The scheduler updates merged for Linux 7.3 include work to help gaming on old hardware, lower latency for short tasks and better handling of Intel’s hybrid processors, Phoronix reported.

Flattening the Pick

The largest change is a patch series from Peter Zijlstra that reworks cgroup scheduling. Phoronix said the series helped gaming performance in testing on “potato” hardware, an Intel Sandy Bridge CPU paired with an AMD Radeon RX 580. The pull request summarized it as a series to “flatten the pick,” improving “cgroup scheduling, which has always been problematic and painful, which has caused various scheduling misbehavior such as the mishandling of reniced tasks et al.”

The series adds several cgroup weight distribution methods through a cgroup_mode setting, with options named up, max, concur and tasks. The default, concur, is described as “the most precise yet also most expensive version.” Cgroup scheduling also moves to a single runqueue.

Latency and Hybrid CPUs

Separately, Vincent Guittot’s changes lower scheduling latency for short-slice tasks, with significant improvements in cyclictest results. The merge also includes a fix to cluster scheduling on CPUs with asymmetric capacity, such as Intel’s hybrid designs, and Linux 7.3 now prefers fully idle cores when balancing load for NOHZ idle CPUs.

Early Benchmarks Are Mixed

Halfway through the merge window, Phoronix ran early Linux 7.3 benchmarks on an HP Z4 G6i workstation with a 48-core, 96-thread Intel Xeon 678X processor, comparing Linux 7.0, 7.1, 7.2 and a 7.3 Git snapshot across dozens of workloads on an Ubuntu 26.10 software stack. The results showed some performance improvements and some regressions, which is typical for a kernel caught mid-merge window, before the release candidate cycle has a chance to fix issues.

For desktop and gaming users, the cgroup changes matter most because modern desktops place applications in cgroups by default, so smarter weight distribution can affect how games compete with background work. Server operators should benchmark their own workloads, since the more precise default mode carries extra overhead.