_picon Posted March 15, 2023 Share Posted March 15, 2023 Hi, I noticed through RenderDoc that the frame rendering starts out roughly with a bunch of ClearUnorderedAccessViewUint calls. They actually take some GPU time, on the order of 0.4ms on a 3060 Ti. But it seems half of them are redundant, so each buffer is cleared twice with the same value. It would make sense to only clear them once, or so I think... Should hopefully be a small/easy fix, with a measurable GPU performance improvement. My settings was "high", and the latest Stable release was used (2.8) Best Regards. 5 4 Link to comment Share on other sites More sharing options...
Flappie Posted April 2, 2023 Share Posted April 2, 2023 Hi @_picon. Can you please show me how to see what you're talking about? 2 Don't accept indie game testing requests from friends in Discord. Ever. Link to comment Share on other sites More sharing options...
_picon Posted April 7, 2023 Author Share Posted April 7, 2023 (edited) I can provide more details if needed, like the entire capture (but it is fairly big!), but this a screenshot of my capture. The "Copy/Clear Pass #1" (expanded) contains a number of ClearUnorderedAccessViewUint calls. If one looks at the resources involved, most are cleared twice, with the same data. (while one is at it, it might be worth looking into if it is necessary to do the clears at all, or if the data is totally overwritten later by compute shaders) Edited April 7, 2023 by _picon (spelling mistake) 1 Link to comment Share on other sites More sharing options...
_picon Posted October 30, 2023 Author Share Posted October 30, 2023 Looks like this inefficiency is still there in 2.9. 1 Link to comment Share on other sites More sharing options...
Flappie Posted November 2, 2023 Share Posted November 2, 2023 I asked devs to take a look. Sorry for the delay. 2 Don't accept indie game testing requests from friends in Discord. Ever. Link to comment Share on other sites More sharing options...
_picon Posted November 10, 2023 Author Share Posted November 10, 2023 Thanks. Perhaps it makes sense to just confirm the same thing doesn't pop up in the new Vulkan based engine. 2 Link to comment Share on other sites More sharing options...
Hiob Posted December 19, 2023 Share Posted December 19, 2023 I just made my usual 4K benchmark. There is a 0.5% overall performance increase, which is of course within margin of error. HOWEVER. The 0.1% fps increased about 25%. Which is great. I don't now if this tweak is responsible. But I just wanted to give some feedback. 17-12-2023, 12:19:12 DCS.exe benchmark completed, 19164 frames rendered in 139.610 s Average framerate : 137.2 FPS Minimum framerate : 124.2 FPS Maximum framerate : 161.8 FPS 1% low framerate : 117.0 FPS 0.1% low framerate : 80.0 FPS 19-12-2023, 19:09:35 DCS.exe benchmark completed, 19278 frames rendered in 139.594 s Average framerate : 138.1 FPS Minimum framerate : 123.1 FPS Maximum framerate : 161.4 FPS 1% low framerate : 119.5 FPS 0.1% low framerate : 101.1 FPS 2 2 "Muß ich denn jedes Mal, wenn ich sauge oder saugblase den Schlauchstecker in die Schlauchnut schieben?" Link to comment Share on other sites More sharing options...
Worrazen Posted December 21, 2023 Share Posted December 21, 2023 (edited) On 11/10/2023 at 8:20 PM, _picon said: Thanks. Perhaps it makes sense to just confirm the same thing doesn't pop up in the new Vulkan based engine. Looks like this was addressed in the update on 19th. Maybe you can test with RenderDoc again to verify the implementation. https://www.digitalcombatsimulator.com/en/news/changelog/openbeta/2.9.2.49629/ Perhaps the issue depends on a particular scene-s, which the benchmark by Hiob might not cover. Then again, small fixes aren't to be shrugged off, they all add up eventually, but it's still a good idea to scrutinize this to determine whether it had the fullest effect. Edited December 21, 2023 by Worrazen Modules: A-10C I/II, F/A-18C, Mig-21Bis, M-2000C, AJS-37, Spitfire LF Mk. IX, P-47, FC3, SC, CA, WW2AP, CE2. Terrains: NTTR, Normandy, Persian Gulf, Syria Link to comment Share on other sites More sharing options...
Hiob Posted December 21, 2023 Share Posted December 21, 2023 @Worrazen My Benchmark scene for sure doesn't cover all aspects. For one, it doesn't feature mirrors or video feeds and no AI for example. However, I don't think that this fix is supposed to make a major impact on average fps. But I saw a significant raise in 0.1% low fps. And this is where (micro-)stuttering e.g. occurs. There are of course a multitude of different reasons possible for such things to improve, but this is exactly the kind of improvement I would expect from such a tweak. And the results are pretty consistent. Of course in the 0.1% lows there is always some variability, but the average over multiple runs is close to a 100, whereas it was ~80 on avg. beforehand. I would consider this a win. 3 "Muß ich denn jedes Mal, wenn ich sauge oder saugblase den Schlauchstecker in die Schlauchnut schieben?" Link to comment Share on other sites More sharing options...
_picon Posted January 2 Author Share Posted January 2 Thanks, will take a look. I've currently having issues with RenderDoc (doesn't attach to the app properly, no API detected) Link to comment Share on other sites More sharing options...
SpectrumZX Posted Tuesday at 09:15 PM Share Posted Tuesday at 09:15 PM В 02.01.2024 в 17:45, _picon сказал: Thanks, will take a look. I've currently having issues with RenderDoc (doesn't attach to the app properly, no API detected) Greetings. Tell me please, did you manage to fix the problem with API detection in the renderdoc? Link to comment Share on other sites More sharing options...
Recommended Posts