Hey people, I am pretty new to CachyOS so bear with me. I installed everything recommended by the Cachy Hello area for gaming including the new GPU Booster for 8GB VRAM cards. My first impressions running a Godot Engine game on Steam “Road to Vostok” were actually really positive, I was getting 30% more performance over Windows both using the same GPU and Vulkan render path. But things changed when I tested a handful of other games that relied on more modern upscalers.
Resident Evil 4 Remake, Deadlock, Marvel Rivals, and Witchfire
FSR 1 performed normally doing it’s spatial scaling and sharpen as expected.
FSR 2 saw framerate get cut in half compared to FSR 1 and the frametimes were worse than that.
XeSS said hold my beer and cut the already disappointing FSR 2 performance and cut it in half again.
TAAU and TSR behaved normally from what I could tell but this solution is heavy compared to FSR.
GPU utilization when it wasn’t spiking was 99% and CPU was typically around 40-50% besides Deadlock which hits CPU harder. When spikes occur neither the CPU or GPU were reading high utilization.
For reference my build is
Operating System: CachyOS Linux
KDE Plasma Version: 6.6.4
KDE Frameworks Version: 6.25.0
Qt Version: 6.11.0
Kernel Version: 6.19.12-1-cachyos (64-bit)
Graphics Platform: Wayland
Processors: 8 × Intel® Core™ i7-6700K CPU @ 4.00GHz
Memory: 32 GiB of RAM (31.3 GiB usable)
Graphics Processor: AMD Radeon RX 5700 XT 8GB
I’m willing to accept that things aren’t going to run perfect on here but this regression seems a little odd, any thoughts would be appreciated.
I have an RX 6600 and from what I can tell, XeSS just doesn’t run well on the early RDNA cards. That said FSR2 and 3 should be fine.
If you’re near the 8gb limit that could be a source of the slowdown. Upscaling techniques use vram themselves. What does your vram usage look like? (steam performance or mango hud can report it)
Yeah I have the steam overlay up all the time to tune things, trying to keep vram around 7gb at most so it has some room to breathe. It could be vram usage but the readouts don’t show anything over budget, in the case of Deadlock the game would have much better frametimes and averages at native 1080p but running FSR2 4K in Performance so 1080p internal runs like garbage.
that’s very odd. I use fsr3 with an rx 580 and it’s helpful so your 5700 xt should be even better I’d think. I don’t know a lot about troubleshooting this particular problem either. Probably the last think I can think of is try opticaler. It also has its own fps output that will show you how much time the game is taking to render and how much time the upscaler is using. It will also come with the latest version which potentially could help but all those games are so new I’d be surprised if there’s any major improvement from that.
Im going to test some more and get some actual data for myself, I’m wondering if this behaviour is related to the GPU being too close to 100% utilization leading to stalls and since FSR runs on the shader cores unlike Nvidia it would actually make the situation worse. Is there a way to check if things like AMD Anti Lag are enabled and a driver level frame limiter?
I disabled the Steam overlay performance monitor and enabled MangoHud configured with Goverlay and now everything works fine, I have no idea what could have caused these spikes but they are gone now?
There is still a weird issue where if I change scalers in game performance takes a nose dive but not in a spikey way, just consistently low fps until the game is restarted. For example in Resident Evil 4 switching from FSR 1 to FSR 2 or vice versa will cause massive slow downs (from the expected average of 60FPS in FSR 2 Quality down to 25-30FPS) until the game is restarted. Then the scaler performs normally which is annoying for testing but not an issue once settings are settled on.