flightlessmango
Better frame pacing for the frame limiter?
I'm really wondering if its possible to add a more accurate frame limiter, i lock the fps to 30 or 60 in some games and even with early or late settings the microstutter persists. RTSS on windows has no such issue and i think its possible to get a better method in place. This really is the only thing stopping me from fully swapping over from windows.
It would depend on how large the micro stutters are. If they're longer than 33ms, then setting it to 30fps would still result in micro stutters
for me mangohud constantly jumps between the frame limit and 1-2fps below it causing games to stutter regardless of how i set the limiter.
Previously DXVK/VKD3D had limtiers that provided smooth FPS but now those are being nuked so mangohud is the ONLY option. In my case the stutters make it unusable for this use case.
I don't have this, I get a stable frames with a variance of about 0.5ms. It would be great to see a screenshot of this. When you say stutter, do you mean the game actually stutters or are you just referring to a spike in the frame time graph? Do you mean the fps limiter causes the game to stutter? Can you post a log file so we can see the frame timings?
I don't have this, I get a stable frames with a variance of about 0.5ms. It would be great to see a screenshot of this. When you say stutter, do you mean the game actually stutters or are you just referring to a spike in the frame time graph? Do you mean the fps limiter causes the game to stutter? Can you post a log file so we can see the frame timings?
My primary game for testing this is Monster Hunter Wilds but unfortunately I'm in Denuvo DRM jail right now from testing proton versions for 24hrs. Wilds suffers the most of my games, with late limiter being plain detrimental to performance (dropping from 80->73fps with 80 limit)
The game actually visibly stutters, I don't even watch the frame time graph/have it displayed. The system is an Alienware Aurora 16x so its an optimus/dual GPU system which complicates getting good presentation even further.
I stopped using mangohud for this purpose previously on my all AMD Tuf A16 Advantage before it died for similar reasons.
I'm currently on Windows again due to getting fed up with trying to make this hybrid system not painful for months but can make a testing Linux install once my Denuvo jail sentence ends.
so after doing some testing on Windows, Monster Hunter Wilds seems to exhibit this behavior there also using RTSS, in Wilds case it looks like Mangohuds frame limiter issues was a symptom not a cause and somehow DXVK limiter did something to resolve it. EDIT: it was reframework on Windows, cant test on Linux yet.
once github git clone starts working again i can get my linux install fully working to do testing on other games. Remnant 2 was another one with issues also, Along with MH Rise and a few others.
Ok so the behavior on Linux for Wilds continues without reframeworkso its still a good test case for this behavior, here is the late method csv log i made. I did my best to trim out the loading screen and main menu so you just see what happens in game and logged to a usb drive so the main drive would have no logging to it.
Late still suffers but not as bad as with reframework, early also does but only dips to 78FPS which i can log if needed. Remnant 2 also exhibits this behavior as does basically any game i own using framegen. Ill get some non framegen games tomorrow also along with Remnant 2 log just for completeness as i do not work tomorrow.
I wonder though, could the fact that Linux is unable to read the EDID of my display properly contribute to this? It doesnt recongize any display mode but the top 240hz mode, that it has VRR, or even the name of the display and Im not 100% confident that top mode is being set right. That begs the question though why didnt DXVK limiter have this issue?
MonsterHunterWilds_trimmed.csv