Benchmark #1512

Download
Elden Ring | BORE vs SCX

CPU Limited scene in Gatefront Ruins (AI heavy) with no background load. 30 second frame recording with 0ms interval. Performance Governor/EPP. SCX commit: 975d65a | Patched LAVD so that Preferred Cores are still identified correctly but not split into two groups since it is detrimental to performance when the difference is minor (due to binning).

Submitted 6 days ago by summedkibbles36

Specifications
Label OS GPU CPU RAM OS specific
BORE Steam Runtime 3 (sniper) AMD Radeon RX 7900 GRE (RADV NAVI31) AMD Ryzen 7 7800X3D 8-Core Processor 33 GB 6.15.0-cachyos1.lto.fc42.x86_64 performance
BPFLAND -m all -w Steam Runtime 3 (sniper) AMD Radeon RX 7900 GRE (RADV NAVI31) AMD Ryzen 7 7800X3D 8-Core Processor 33 GB 6.15.0-cachyos1.lto.fc42.x86_64 performance
BPFLAND -m all -w --slice-us 1000 Steam Runtime 3 (sniper) AMD Radeon RX 7900 GRE (RADV NAVI31) AMD Ryzen 7 7800X3D 8-Core Processor 33 GB 6.15.0-cachyos1.lto.fc42.x86_64 performance
Patched LAVD --performance Steam Runtime 3 (sniper) AMD Radeon RX 7900 GRE (RADV NAVI31) AMD Ryzen 7 7800X3D 8-Core Processor 33 GB 6.15.0-cachyos1.lto.fc42.x86_64 performance
Patched LAVD --performance --slice-max-us 1000 Steam Runtime 3 (sniper) AMD Radeon RX 7900 GRE (RADV NAVI31) AMD Ryzen 7 7800X3D 8-Core Processor 33 GB 6.15.0-cachyos1.lto.fc42.x86_64 performance
# Top runs: * **Highest FPS**: `BPFLAND -m all -w` with approximately 5% higher average FPS compared to others. * **Smoothest FPS**: `Patched LAVD --performance --slice-max-us 1000` with the lowest standard deviation and variance in FPS. * **Best overall**: `BPFLAND -m all -w --slice-us 1000` with performance slightly lower than the highest FPS run but significantly more stable and smoother. # Issues: * The patched `LAVD --performance` scheduler demonstrated approximately 7% lower average FPS than the top-performing scheduler implementations. This indicates inefficiency in handling the CPU-limited scenario. * Another run, `BORE`, exhibited around 3% lower FPS and higher frame time variability compared to the top-performing runs, indicating a drop in performance and smoothness. # Summary This benchmark compares various `SCX` scheduler implementations against the `BORE` scheduler and highlights significant differences in performance during a CPU-limited scenario in Elden Ring. The `BPFLAND -m all -w` scheduler excelled, offering the highest FPS, whereas `BPFLAND -m all -w --slice-us 1000` provided the most balanced performance profile, combining both high FPS and smoothness. The `LAVD` related patched configurations showed improvements but still lagged behind in FPS, indicating room for optimizations. These insights help depict how different schedulers can affect game performance, particularly in demanding scenes like Gatefront Ruins.