doitsujin
Call of Duty: MWR - Graphic glitches during mission "blackout"
Software information
Call of Duty - Modern Warfare Remastered Maximum settings except a few things. I reckon the glitch goes away if I lower the quality to minimum.
System information
- GPU: Nvidia GTX 1050TI
- Driver: 520.56.06
- Wine version: 7.20
- DXVK version: 2.0 - Release
Apitrace file(s)
Not created, yet.
Log files
h1_sp64_ship_dxgi.log h1_sp64_ship_d3d11.log
info: d3d11.samplerAnisotropy = 16
Try it without that option. It can definitely break games.
If that doesn't work, please make an apitrace.
info: d3d11.samplerAnisotropy = 16
Try it without that option. It can definitely break games.
If that doesn't work, please make an apitrace.
I'll try it EDIT: It doesn't fix anything
If that doesn't work, please make an apitrace.
With DXVK the game is unbearably slow, it's even hard to see any glitch. Without DXVK "directX error" and I cannot apitrace anything.
With DXVK the game is unbearably slow, it's even hard to see any glitch.
Doesn't matter, we need the apitrace.
With DXVK the game is unbearably slow, it's even hard to see any glitch.
Doesn't matter, we need the apitrace.
Even if it's with DXVK? Ok then, let's do it.
I'm about to try and record another apitrace, this exact type of glitches are present in other levels. This level is faster to load, less useless frames will be captured and it's more "downloadable" for you, and easier to upload for my capped data.
The shadows are the problem. If I turn them off the glitches disappear. Why?
Can you post some screenshots? The apitrace looks fine.
Yeah, the glitch disappear when I record the apitrace, that's just a bad joke. Anyway, I have to try with a recent version of the game, I guess it's an issue with the game.
Adding to the issue that i can reproduce it ingame. Haven't found it to be a regression as it was the same with dxvk 1.0. The issue appears when you have shadows turned on plus "CACHE SUN SHADOW MAPS" turned off. If you turn off shadows it seems to go away and if you have shadows on then turning on "CACHE SUN SHADOW MAPS" also seemed to make it go away.
This is on a R9 380
turning on "CACHE SUN SHADOW MAPS" also seemed to make it go away.
Jesus, that option will just use so much VRAM that'll kill my smooth performance... Anyway, regression, how? What's exactly the culprit, DXVK-wise, I mean.
Anyway, regression, how?
No i meant it isn't a dxvk regression as it appeared in even the oldest dxvk versions i tried.
What's exactly the culprit, DXVK-wise, I mean.
No idea as i have no technical knowledge about this and just try to help testing :grin: Also when it's buggy in this game section it has a tendency to freeze my entire graphical output so i had to force down my pc quite alot when testing this lol.
Anyway, regression, how?
No i meant it isn't a dxvk regression as it appeared in even the oldest dxvk versions i tried.
What's exactly the culprit, DXVK-wise, I mean.
No idea as i have no technical knowledge about this and just try to help testing grin Also when it's buggy in this game section it has a tendency to freeze my entire graphical output so i had to force down my pc quite alot when testing this lol.
I enabled the other cache thing too, I had to lower the shadows to "medium" because with DXVK you need to leave some room...
EDIT: Nevermind, I'd gladly see this fixed somehow, as the shadow spot precache tanks the FPS by a lot.
Yes i have reproduced this issue on my R9 380. Tho it is a bit annoying since it also has a tendency to hang my GPU so had to force reboot quite a few times. I don't have any technical knowledge to begin digging deeper into it but I've tried various config options and old versions without luck so far. Will also try to check it out on the Nvidia setup or Steam Deck.
@K0bin I've managed to trace more glitches, while trying to replay the trace I noticed strange stuff:
- The frames with the glitches are being completely skipped
- The last frame has a gigantic number of draw-calls, like 20 frames were being combined, even more.
I'm uploading the trace now, meanwhile, could you please explain this to me?
EDIT: I forgot to mention that the latest nvidia driver (525.85.05) has corrected just one "terrain glitch", the one that I was showing in the video on the first comment.
https://mega.nz/file/Q8MmUAzT#zwXUJHzkko1RRhz62O1Vm1Un0vgPB10DqiGd0m413hA
Your trace crashes on replay but i managed to get one myself after playing at 0.4fps for half an hour. Sadly it doesn't reproduce with it. Renderdoc also makes the game crash on start. Only clue i've found is that it seems to be a old regression in https://github.com/doitsujin/dxvk/commit/8b5db80fbd3151084e5efbf32933cae13d5e399d
How did you get this info?
On Sat, 11 Feb 2023, 18:21 Blisto91, @.***> wrote:
Your trace crashes on replay but i managed to get one myself after playing at 0.4fps for half an hour. Sadly it doesn't reproduce with it. Renderdoc also makes the game crash on start. Only clue i've found is that it seems to be a old regression in 8b5db80 https://github.com/doitsujin/dxvk/commit/8b5db80fbd3151084e5efbf32933cae13d5e399d
— Reply to this email directly, view it on GitHub https://github.com/doitsujin/dxvk/issues/3073#issuecomment-1426830626, or unsubscribe https://github.com/notifications/unsubscribe-auth/AC47ERND3FWQBEEYAPHEG4DWW7DAXANCNFSM6AAAAAASCOLYDY . You are receiving this because you authored the thread.Message ID: @.***>
Slow and painful testing (mostly because it has a tendency to hang my GPU)
Yeah but did you inspect the trace to find it?
On Sat, 11 Feb 2023, 19:17 Blisto91, @.***> wrote:
Slow and painful testing (mostly because it has a tendency to hang my GPU)
— Reply to this email directly, view it on GitHub https://github.com/doitsujin/dxvk/issues/3073#issuecomment-1426843955, or unsubscribe https://github.com/notifications/unsubscribe-auth/AC47ERKXPO4VKA6J2EOAOATWW7JU5ANCNFSM6AAAAAASCOLYDY . You are receiving this because you authored the thread.Message ID: @.***>
Slow and painful testing (mostly because it has a tendency to hang my GPU)
You know what's really sad? The only two games that gives me real problems are almost impossible to debug, can't trace them, can't view what's been traced because "yes"...
Heh, nice find by the way! @doitsujin it is possible to try to revert that commit for testing if it solves the problem with those weird two remaster? I hope there's a switch or something in dxvk.conf though.
@doitsujin it is possible to try to revert that commit for testing if it solves the problem with those weird two remaster? I hope there's a switch or something in dxvk.conf though.
Sadly not possible as we don't know what the issue is yet. The commit isn't small and touch alot of d3d11 (also it's ancient)
Only clue i've found is that it seems to be a old regression in 8b5db80
How did you narrow the issue down to, possibly, this commit?
By gong back to very old versions and noticing it wasn't a issue there. And then afterwards bisect the specific commit it appeared
@gabriele2000 Are you able to verify that this works with dxvk master?
Once I'll reinstall the two games, I'll let you know. It won't be long until I do, maybe less than one month.
@gabriele2000 Friendly ping
I'll install the game soon, after the 24th, before the end of the month. Reason: internet cap and I had to upload a ton of stuff those months.