NV Reflex badly render submit application, compared with force dll injected (SK)

[angelooo18]

Hi Dev Team,

I hope this email finds you well.

I wanted to take a moment to express my sincere gratitude for your incredible work on CS. I've been playing for years, and it has brought me so many good memories. Thank you for creating such an enjoyable and memorable experience.

Now going to the point, from some time ago I began testing with NSIGHT and XTW, the performance of the game within GSYNC-VSYNC and Reflex. The thing is that apparently reflex seems to have a heavy frametime stuttering problem regardless of the configuration used. This indicated that the frame pacing is really bad at this point and only some external tools can fixed (maybe deriving into an VAC ban) So to test this out, I run a default test in a Windows 10 IoT LTSC 21H2 19044.1288 with current Windows Updates and the latest available version of the NVIDIA driver at the time of this mail. (6700k, 32GB JEDEC DDR4 noXMP, 500GB m.2, GTX1080 11Gbps) So here I attach the analysis and files from the test, please find them useful as this was the average from 3 tests, restart each time and filtered out. MsCPUWait, MsUntilDisplayed and Estimated PCL was superb with SpecialK meaning that the force of the NVAPIs for Reflex are working better in that regard, Simulation and Presents becomes really stable in comparison as native reflex have highly variation (form 5ms to 9ms randomly without any input), I hope you guys can find a gay to check on this as is incredible how an third party tool can perform better on the analysis

Best regards, Angelo Lorenzoni

Files to check and compare (ETL are not shared as they contain sensetive information) Only summaries:

Native Reflex Summary.csv Native Reflex.csv Native Reflex.txt SpecialK Summary.csv SpecialK.csv SpecialK.txt