Benchmark #1375

Download
Risk of Rain 2 | BORE vs SCX

CPU Limited scene from the first stage with a stable number of enemies, no background load. 30 second frame recording with 0ms interval. All schedulers were running in performance mode except LAVD --autopilot. Performance Governor/EPP

Submitted 3 weeks ago by summedkibbles36

Specifications
Label OS GPU CPU RAM OS specific
BORE Freedesktop SDK 24.08 (Flatpak runtime) AMD Radeon RX 7900 GRE (RADV NAVI31) AMD Ryzen 7 7800X3D 8-Core Processor 33 GB 6.14.4-cachyos2.lto.fc43.x86_64 performance
BPFLAND --local-pcpu Freedesktop SDK 24.08 (Flatpak runtime) AMD Radeon RX 7900 GRE (RADV NAVI31) AMD Ryzen 7 7800X3D 8-Core Processor 33 GB 6.14.4-cachyos2.lto.fc43.x86_64 performance
BPFLAND --local-pcpu --no-wake-sync Freedesktop SDK 24.08 (Flatpak runtime) AMD Radeon RX 7900 GRE (RADV NAVI31) AMD Ryzen 7 7800X3D 8-Core Processor 33 GB 6.14.4-cachyos2.lto.fc43.x86_64 performance
BPFLAND -m performance Freedesktop SDK 24.08 (Flatpak runtime) AMD Radeon RX 7900 GRE (RADV NAVI31) AMD Ryzen 7 7800X3D 8-Core Processor 33 GB 6.14.4-cachyos2.lto.fc43.x86_64 performance
BPFLAND --no-preempt Freedesktop SDK 24.08 (Flatpak runtime) AMD Radeon RX 7900 GRE (RADV NAVI31) AMD Ryzen 7 7800X3D 8-Core Processor 33 GB 6.14.4-cachyos2.lto.fc43.x86_64 performance
BPFLAND --no-wake-sync Freedesktop SDK 24.08 (Flatpak runtime) AMD Radeon RX 7900 GRE (RADV NAVI31) AMD Ryzen 7 7800X3D 8-Core Processor 33 GB 6.14.4-cachyos2.lto.fc43.x86_64 performance
BPFLAND --no-wake-sync --no-preempt Freedesktop SDK 24.08 (Flatpak runtime) AMD Radeon RX 7900 GRE (RADV NAVI31) AMD Ryzen 7 7800X3D 8-Core Processor 33 GB 6.14.4-cachyos2.lto.fc43.x86_64 performance
LAVD --autopilot Freedesktop SDK 24.08 (Flatpak runtime) AMD Radeon RX 7900 GRE (RADV NAVI31) AMD Ryzen 7 7800X3D 8-Core Processor 33 GB 6.14.4-cachyos2.lto.fc43.x86_64 performance
LAVD --no-preemption Freedesktop SDK 24.08 (Flatpak runtime) AMD Radeon RX 7900 GRE (RADV NAVI31) AMD Ryzen 7 7800X3D 8-Core Processor 33 GB 6.14.4-cachyos2.lto.fc43.x86_64 performance
LAVD --no-wake-sync Freedesktop SDK 24.08 (Flatpak runtime) AMD Radeon RX 7900 GRE (RADV NAVI31) AMD Ryzen 7 7800X3D 8-Core Processor 33 GB 6.14.4-cachyos2.lto.fc43.x86_64 performance
LAVD --no-wake-sync --no-preemption Freedesktop SDK 24.08 (Flatpak runtime) AMD Radeon RX 7900 GRE (RADV NAVI31) AMD Ryzen 7 7800X3D 8-Core Processor 33 GB 6.14.4-cachyos2.lto.fc43.x86_64 performance
LAVD --performance Freedesktop SDK 24.08 (Flatpak runtime) AMD Radeon RX 7900 GRE (RADV NAVI31) AMD Ryzen 7 7800X3D 8-Core Processor 33 GB 6.14.4-cachyos2.lto.fc43.x86_64 performance
# Top runs: * **Highest FPS**: `BORE` has the highest average FPS, providing about 14% more frames per second compared to the second-best run. * **Smoothest FPS**: `BORE` also shows the smoothest performance with the lowest standard deviation and variance in FPS, reflecting consistent frame delivery. * **Best overall**: `BORE` stands out as the best overall due to both highest performance and consistent frame times. The balance between FPS and stability makes it superior over other runs. # Issues: * Several runs under the `BPFLAND` label, particularly those using `--no-preempt` and `--no-wake-sync --no-preempt`, yielded noticeably lower FPS, by approximately 25% compared to the better performing runs. The significantly lower FPS in these runs could hint at configuration or scheduling inefficiencies when certain flags are used. * The `BPFLAND --no-preempt` and other similar configurations experienced increased variability in FPS, contributing to less stable gameplay experience. # Summary In this benchmark comparing different schedulers and configurations in a CPU-constrained scene in *Risk of Rain 2*, the results show that `BORE` outperforms its counterparts significantly, with superior FPS and stability. The `BPFLAND` runs with `--no-preempt` and similar flags show subpar performance and increased variability, suggesting inefficiencies in those configurations. Optimization seems critical for such scheduling strategies to potentially reach the performance levels of `BORE`. The test environment remained consistent across all runs, ensuring that observed differences are attributable to the scheduler configurations examined rather than hardware discrepancies.