I'm just not even going to engage with this nonsense.
Either bisect to a specific commit or limit your FPS or whatever, or just stick to 2.7.1 if you "prefer" lower perf or can't figure out compile options. What exactly are you expecting us to do here without even mentioning an affected game?
I was going to do this investigation myself.
I asked if you can see something on your side with all your usual frameworks and tools already set up.
I created this issue to track changes and to share this with you.
Me, please remind me to not deal with dxvk/valve devs any more.
No offence, but 90% of your reports and comments are monologues or are complete non-issues for the vast majority of users anyway.
If you think there's higher CPU utilization then give us a concrete example that's easy to reproduce, ideally with the CI builds because not everyone here uses Gentoo. Anything else has no value.
@pchome
and
may be the reason why are you seeing higher CPU usage in percentage.
DXVK now uses less CPU threads in certain situations by default, and in monitoring software it may look like higher CPU usage in percentage, due to monitoring software logic.
d3d11.dllx1 2026-06
Setup
I was using DXVK 2.7.1 (stable Gentoo build).
Now for testing purposes I built DXVK 3.0 (
emerge -1B dxvk -q).Usage
I unpacked binary package (that's why -B flag) and used files on some random games.
Since I prefer low CPU usage in games (many can run at <=10%)[^1] I noticed higher CPU usage with DXVK 3.0.
When CPU usage previously was ~ 1?%[^2] now 2?% with DXVK 3.0.
Debug
Quick
perfdebug show nothing helpful:libvulkan_radeon.soin my case).d3d11.dllandntdll.so(like half of all load)Since the build have no symbols so no details available.
I'll try to build WINE and DXVK with debug info to investigate further, but maybe there is something developers can see/do before I'll try to reintroduce myself to all this mingw/crosscompiling debugging.
[^1]: Especially when it's hot outside. To not trigger CPU fan too often.
[^2]: Replaced
1xwith1?to avoid confusion, mean 10-19.