Unified back/depth buffer - 0x00000001 OGL_SINGLE_BACKDEPTH_BUFFER_ENABLE Memory Allocation Policy - 0x00000002 WKS_MEMORY_ALLOCATION_POLICY_AGGRESSIVE_PRE_ALLOCATION Power management mode - 0x00000003 PREFERRED_PSTATE_PREFER_CONSISTENT_PERFORMANCEįrame Rate Monitor - 0x4A5A3219 PS_FRAMERATE_LIMITER_GPS_CTRL_OPTIMAL_SETTINGįrame Rate Monitor Control - 0x00000010 PS_FRAMERATE_MONITOR_CTRL_FRL_OFFSET_SHIFT These settings are optimized for the latest drivers and can deliver significant improvements in Source Engine and Goldsrc games.įrame Rate Limiter Mode - 0x00000010 PS_FRAMERATE_LIMITER_2_CONTROL_ALLOW_ALL_MAXWELL Anyways when you post your findings on the reg key could you also include your windows version as well as nvidia driver version.Ĭode: Select all Open up NVIDIA Inspector and show unknown settings (gears icon on top right).Īpply these settings ON A GAME PROFILE. Its interesting that you did not notice any input improvement. All the tests above were done with 'DisableWriteCombining' on a win7, Nvidia 441.41 driver, 3700x, 1050ti, 720p, 3200mhz 16-18-18-36 system.įrom my experience, in terms of input (tested in apex, kovaaks, diabotical, quake live) Riot LLHP > Melody Mod > Base. The base profile is everything default expect putting power management to 'max performance', max pre-rendered frames to '1' and other basic tweaks you find on every yt videos.

I currently use the 'Riot LLHP' profile since input was better on it despite the lower 1% and 0.1% lows compared to 'Melody Mod' profile. Also as I noted on my OP, these profiles might cause massive fps drops on OpenGL or Vulkan games. The fps reduction is somewhat game dependent I believe, since from my benchmarks I didn't lose any fps on Apex Legends, infact 1% and 0.1% lows were better with the melody profile.

It's not a big deal since I use separate OS for CS:GO and still have high FPS during actual match, but I didn't notice any improvements in responsiveness.

FPS in CS:GO benchmark map was also reduced from 485-490 to 460. I've tried 'Riot Low_Latency_High_Performance v2' profile and it reduced 'Video Memory Throughput' from 98-99k to 67k in winsat formal results.

Nvidia inspector tweaks