Performance on Black Myth Wukong with 6800 XT and 5800X3D
Performance on Black Myth Wukong with 6800 XT and 5800X3D
I’ve conducted some quick tests with their open benchmark tool to evaluate how the "game" functions. The test runs smoothly on Linux, though I have no certainty about the actual application. On Fedora40 with a Ryzen 7 5800X3D and an RX 6800 XT, running at 2600MHz with manual tuning, I applied a small optimization to the GPU. The GPU has a 50mV core voltage and a power cap of 312W. Results showed: Motion Blur is absent in all 1440p high resolutions. Native AVG was 65 FPS, max 75 FPS, min 55 FPS, low 5th 58 FPS. At 1440p high with 75% FSR scaling, average was 87 FPS, max 106 FPS, min 73 FPS, low 5th 78 FPS. 1440p high with 75% TSR gave 85 FPS (max 100), min 70 FPS, low 5th 76 FPS. In still scenes, FSR outperformed in visuals despite slight blur. For static frames, FSR clearly shines. Running most settings at Very High should work around 60 FPS with upscaling. The benchmark itself is very GPU-intensive and puts minimal strain on the CPU—my 5800X3D used under 20W for the whole session. It’s possible the real game could be more CPU demanding than this. You can likely enable Cinematic textures without major performance drops, provided your card has sufficient VRAM; textures look great in that mode. Also worth noting: the game consistently uses Lumen on DX12, so disabling it on Windows requires running it on DX11. I suspect the benchmark didn’t run Lumen on Linux via Proton when using Vulkan.
I've tested several runs and performed well. Of course I'm using 1080p and a 6800, but with RTX on low and the rest settings at 70% FSR upscaling, I'm achieving around 56fps. That's a decent frame rate for a game like this. Surprisingly, it's more than enough for playability, especially compared to what I expected. It's made me reconsider whether I should try the game again, as I had hoped it wouldn't work on my current setup. Still thinking about how my old PC could handle it.
It's also this benchmark tool that works well, yet it feels oddly missing the combat elements of the game. It's strange that none of the visual effects like sparks or graphics interactions would show up in a graphics test. Keep in mind, ray tracing performed nicely, but when turned on I noticed an unusual blue light effect. I'm not sure if the cause lies in the game, the hardware, or the software.
It appears RT isn't functioning properly on Linux, especially when maxed out. FPS remains steady and visuals don’t improve. Blue light seems irrelevant. After restarting the program, performance dropped to 10–20 FPS even at low settings with FSR at 75%. Even at reduced settings, it jumps between 10–20 FPS on 1440p. Frame generation still works, yielding around 100–110 FPS at full native resolution.
I also saw the RT medium appears quite poor. There were some sparkly elements in the grass at one spot. It seems like it's based on low ray count rather than bounce count, which results in low resolution after denoising. There’s a discussion thread about the benchmark that’s hard to find a clear title for. I’ve shared my notes there already.
On my 6950xt I achieve a solid 16:9 1440p 60fps setup with Shadows, Visual Effect, Hair/Vegetation/Global Illumination at High and 75% resolution. Reducing these settings noticeably degrades visual quality in my view, though I can tolerate lower values for Effects and Global Illumination if necessary. At 21:9 I must significantly cut graphics for personal preference, so I settle for 30 fps with maximum settings. Likely not ideal at that frame rate, but the benchmark still looks decent.
Sure, review your complete comment in the referenced discussion for more background if needed.