http://bugs.winehq.org/show_bug.cgi?id=10400
Summary: Horible performance with OffscreenRenderingMode=fbo in Counter Strike source "Video stess test" Product: Wine Version: 0.9.49. Platform: PC URL: http://appdb.winehq.org/objectManager.php?sClass=version &iId=3731 OS/Version: Linux Status: NEW Severity: major Priority: P2 Component: wine-directx-d3d AssignedTo: wine-bugs@winehq.org ReportedBy: vitaliy@kievinfo.com
With OffscreenRenderingMode set to "fbo" I'm getting less then 1 fps in Counter Strike Source Video stress test. Disabling monitors with "cl_drawmonitors 0" makes it much faster (30+ FPS).
Major because it affects lots of other games. Also since development goes towards making FBO a default instead of back buffer.
http://bugs.winehq.org/show_bug.cgi?id=10400
--- Comment #1 from Vitaliy Margolen vitaliy@kievinfo.com 2007-11-10 21:38:25 --- All other d3d settings are by default. My hardware is: Video : nVidia 6600GT 128MB Driver: nVidia 100.14.19
http://bugs.winehq.org/show_bug.cgi?id=10400
Vitaliy Margolen vitaliy@kievinfo.com changed:
What |Removed |Added ---------------------------------------------------------------------------- Summary|Horible performance with |Horrible performance with |OffscreenRenderingMode=fbo |OffscreenRenderingMode=fbo |in Counter Strike source |in Counter Strike source |"Video stess test" |"Video stress test"
http://bugs.winehq.org/show_bug.cgi?id=10400
--- Comment #2 from H. Verbeet hverbeet@gmail.com 2007-11-11 16:43:37 --- FWIW, this appears to be hardware or setup specific. Since I'm unable to reproduce the issue over here (GeForce Go 7900 GS 256MB / 100.14.23) it'll be pretty hard to even start fixing this, for me at least. Perhaps Stefan or Roderick are able to reproduce this?
http://bugs.winehq.org/show_bug.cgi?id=10400
--- Comment #3 from Vitaliy Margolen vitaliy@kievinfo.com 2007-11-11 18:35:19 --- So which part of the multi-gigabyte logs that Wine can produce can answer that question? I tried to reset all the video configuration I could find. And remove any HW cards that might have anything to do with this. Still no go.
AFAIK FBOs should work with my card and caps show that.
http://bugs.winehq.org/show_bug.cgi?id=10400
--- Comment #4 from H. Verbeet hverbeet@gmail.com 2007-11-12 11:21:53 --- (In reply to comment #3)
So which part of the multi-gigabyte logs that Wine can produce can answer that question?
Logs probably won't help much here since it's a performance issue, but if there's anything to see in the logs likely places would be anything generated inside apply_fbo_state() or depth_copy() (you could try disabling depth_copy() completely, although I doubt it'll make much of a difference). In general you'd want to narrow things down to a specific FBO configuration that causes the slowdown like a specific texture format or active render states.
I tried to reset all the video configuration I could find. And remove any HW cards that might have anything to do with this. Still no go.
AFAIK FBOs should work with my card and caps show that.
A 6600GT should support FBOs just fine. The only other difference with my setup that I see that could be significant is the amount of video memory.
http://bugs.winehq.org/show_bug.cgi?id=10400
--- Comment #5 from Vitaliy Margolen vitaliy@kievinfo.com 2007-11-18 15:06:47 --- It might be something in my setup/configuration. The question is what? Trying different things some how I've managed to get good FPS with fbo in cs:s. One of the things that did some difference was AGP aperture size that was changed from 128MB to 256MB.
However it still does not work always. What makes it work/not work I have no clue. Is there an easy way to see what's different as far as Wine sees?
Not closing it yet, because it's too intermittent to be sure which is it.
http://bugs.winehq.org/show_bug.cgi?id=10400
--- Comment #6 from Vitaliy Margolen vitaliy@kievinfo.com 2007-11-29 22:11:29 --- The only additional piece of information I have is this message printed 100 times for each frame: warn:d3d_shader:IWineD3DVertexShaderImpl_CompileShader Swizzled attribute validation required an expensive comparison
http://bugs.winehq.org/show_bug.cgi?id=10400
--- Comment #7 from H. Verbeet hverbeet@gmail.com 2007-11-30 02:57:15 --- Is that with FBOs only? That warning was added quite recently afaik. I guess enabling the FBO path might trigger different shaders or vertex declarations to be used, but I don't think we report much different caps based on the ORM being used.
http://bugs.winehq.org/show_bug.cgi?id=10400
--- Comment #8 from Vitaliy Margolen vitaliy@kievinfo.com 2007-11-30 07:24:58 --- You right, it's not just with fbo. With default off screen rendering I'm getting this warning as well.
http://bugs.winehq.org/show_bug.cgi?id=10400
--- Comment #9 from Vitaliy Margolen vitaliy@kievinfo.com 2007-12-15 11:16:00 --- Still present in 0.9.51. Setting "cl_drawmonitors 0" immediately improves performance to somewhere past 35 FPS. Setting "cl_drawmonitors 1" slows it down to a crawl again.
http://bugs.winehq.org/show_bug.cgi?id=10400
--- Comment #10 from Vitaliy Margolen vitaliy@kievinfo.com 2008-01-30 20:25:08 --- With nVidia's 169.09 drivers it's much better. However now rendering "pauses" each 2 sec in the first portion of the stress test (rotating panels). Again regardless of the dxlevel.
http://bugs.winehq.org/show_bug.cgi?id=10400
--- Comment #11 from Vitaliy Margolen vitaliy@kievinfo.com 2008-03-01 21:51:40 --- With nVidia 169.12 the problem is back. So whatever they are fixing for HL1 games affect fbo.
http://bugs.winehq.org/show_bug.cgi?id=10400
Vitaliy Margolen vitaliy@kievinfo.com changed:
What |Removed |Added ---------------------------------------------------------------------------- Status|NEW |RESOLVED Resolution| |INVALID
--- Comment #12 from Vitaliy Margolen vitaliy@kievinfo.com 2008-03-01 21:57:17 --- I take that back. It was the major slowdown without fbo! With fbo it's now same fast as it used to be.
Closing invalid.
http://bugs.winehq.org/show_bug.cgi?id=10400
Vitaliy Margolen vitaliy@kievinfo.com changed:
What |Removed |Added ---------------------------------------------------------------------------- Status|RESOLVED |CLOSED
--- Comment #13 from Vitaliy Margolen vitaliy@kievinfo.com 2008-03-01 21:57:34 --- Closing.