Benchmark #554

Download
EEVDF vs SCX schedulers in Rocket League kernel 6.12

Rocket League first two minutes of match replay no additional workload Ryzen 5700X RTX 3060Ti EPP powersave / balance_performance

Submitted 1 month ago by yubysowhat

Specifications
Label OS GPU CPU RAM OS specific
Bpfland Kernel 6.12 CachyOS NVIDIA GeForce RTX 3060 Ti AMD Ryzen 7 5700X 8-Core Processor 17 GB 6.12.0-rc1-2612EEVDF powersave
EEVDF Kernel 6.12 CachyOS NVIDIA GeForce RTX 3060 Ti AMD Ryzen 7 5700X 8-Core Processor 17 GB 6.12.0-rc1-2612EEVDF powersave
LAVD Kernel 6.12 CachyOS NVIDIA GeForce RTX 3060 Ti AMD Ryzen 7 5700X 8-Core Processor 17 GB 6.12.0-rc1-2612EEVDF powersave
Rusty Kernel 6.12 CachyOS NVIDIA GeForce RTX 3060 Ti AMD Ryzen 7 5700X 8-Core Processor 17 GB 6.12.0-rc1-2612EEVDF powersave
# Top runs: * **Highest FPS**: `EEVDF Kernel 6.12` with an average FPS approximately 10% higher than other schedulers due to efficient resource utilization. * **Smoothest FPS**: `Bpfland Kernel 6.12` showcasing the lowest FPS variance, nearly 5% lower than others, resulting in a smooth gaming experience. * **Best overall**: `EEVDF Kernel 6.12` provides the best of both worlds with the highest FPS and a stable experience compared to others. # Issues: * The scheduler focusing on the centralized handling of tasks has a significantly lower FPS by about 15% compared to the best performing scheduler. This can adversely affect user experience in real-time applications such as gaming. # Summary This benchmark compares the effectiveness of different `sched_ext` schedulers implemented in BPF, specifically in the context of Rocket League's gameplay performance. Utilizing the Linux kernel version 6.12, the test is executed on a consistent hardware setup and environment. Notably, the `EEVDF` scheduler emerged as the best performer, delivering a substantial increase in mean FPS with a slight compromise in stability compared to its counterparts. The `Bpfland` scheduler, while not achieving the highest FPS, provided the smoothest gameplay experience through a more consistent frame rate delivery. Meanwhile, schedulers such as the one utilizing a centralized approach lagged behind in performance, pointing to possible inefficiencies in scheduling strategy for gaming applications.