Benchmark #1094

Download
osu!stable + stress-ng - BORE vs lavd vs bpfland

Submitted 2 weeks ago by .dnaim

Specifications
Label OS GPU CPU RAM OS specific
bore Steam Runtime 3 (sniper) AMD Radeon Graphics AMD Ryzen 5 6600HS Creator Edition 14 GB 6.14.0-rc3-1-cachyos-rc powersave
bpfland Steam Runtime 3 (sniper) AMD Radeon Graphics AMD Ryzen 5 6600HS Creator Edition 14 GB 6.14.0-rc3-1-cachyos-rc powersave
lavd Steam Runtime 3 (sniper) AMD Radeon Graphics AMD Ryzen 5 6600HS Creator Edition 14 GB 6.14.0-rc3-1-cachyos-rc powersave
# Top runs: * **Highest FPS**: `bore` with a mean FPS approximately 4% higher than the baseline due to higher GPU loads and core clock speeds. * **Smoothest FPS**: `lavd` with the lowest standard deviation of FPS, approximately 25% more stable than the baseline, indicating more consistent performance. * **Best overall**: `lavd` offers a good combination of high FPS and the highest stability among the runs, making it the most balanced performer with significantly more stable FPS at only about 3% lower FPS compared to the top FPS run. # Issues: * The `bpfland` run shows a significantly lower average FPS, approximately 33% less than the baseline, with similar variance across other metrics. * All runs used consistent hardware and software configurations, ensuring that the performance discrepancies are purely due to the benchmarks being tested. # Summary This benchmark compared three different schedulers for the osu!stable running alongside `stress-ng` in Linux environments. Specifically, it assessed `bore`, `lavd`, and `bpfland` schedulers. The `lavd` scheduler provided the best balance of high average FPS and minimal frame time variance, making it the superior option for stable and reliable performance. Meanwhile, the `bpfland` and `bore` runs highlighted the impact of scheduling on performance, with significant variation in both FPS and load management. Overall, this test highlights and underscores how scheduler choices on an identical system setup can dramatically impact gaming performance.