NVIDIA Reflex Low Latency - How It Works & Why You Want To Use It

can i just keep the VSycn off from Nvidia cantrol panel when using Nvidia Reflex low latency ?, or keep the VSync Use the 3D application settings for the lowest input lag ?

Is the old method still available in the nvidia drivers that works with any game?

Great video man. Keep up the good work!

thought provoking. easily explained. soothing voice. just the way god made ya :P

Hey man, do you have any idea why AMD anti-lag causes AoE II: Definitive Edition to crash when connecting to MP match? It's very strange.

does it affect fps if i have low latency mode turned on with vsync on as well?

Can you test the new nvidia reflex mode in world of warcraft?

гений открытий, просто человек давший пищу для размышления компании nvidia

So technically this should be set to off if you have a gpu that is anything under 99% usage unless you are playing cod or something? Does it do anything to controllers?

I saw your video recently, what is best for me? low latency mode off, on or ultra? competitively. I play the new cod modern warfare 2. I'm using an i3 9100f and a gtx 1650. In the game I have the nvdia reflex option for on, off or on + boost

Ok Reflex ingame give me more fps at the cost of gpu usage (on my system it almost doubles to like 300 fps) but why i would want more fps than my monitor is capable of refreshing ? I prefer to limit to 144 -4 and have a gpu /cpu stable and cooler in my case of course (144mhz)

best content on the subjet i wish you keep making some more videos

so low latency mode should be set to OFF, if your GPU isn't maxed out?

What should I use on Valorant for a RTX 2060 super? on or on + boost? also does the reflex override my nvidia control panel low latency settings?

I should make it ultra from nvidia or game settings or together?

With an RTX 3090, this guide is golden.

4K 72Hz monitor and some titles just had insane latency that felt horrible while others did not. Using reflex and ULL where possible and a 69Hz FPS cap for everything else just works... everything is smooth.

It feels more or less the same as my 165Hz panel despite the lower frame rates, because the system and render latency end up more or less the same... not much outside of Esports titles on low settings can maintain 160FPS without ever dipping.

In terms of frametimes, should i just lock with RTSS to make it smooth, or would frametimes be good enough with high (200+) framerates + Reflex?

Thanks Chris

Thank you 😊👍

Are there any improvements when you cap FPS and use On + Boost? The graph in the video shows FPS cap with Reflex On, but doesn't include FPS cap with On + Boost.

THis is a paper worth research and effort fitted in a video. Thank you very much.

So when I have 144hz gsync.
I turn on Vsync + ultra low latency + fps cap 141 + nvidia reflex, so is it wrong or good?

Really appreciate the work you've done. Helps me understand this very well. Cheers!

If i have in nvidia settings low latency mode on ultra should I use in games like tarkov, or typical fps games setting in game on plus boost as well ?

Should I be turning OFF Reduce Buffering now?

It's really funny how younger people and corporations find this stuff "new" when it's been this way since the early 90s haha, many people just couldn't be bothered to go and optimize their systems and settings for low latency when they didn't even notice the delays. Where as those who cared and noticed already have this stuff minimized for decades. Setup a framerate limiter as close to the game engine as possible, done, never let FPS go below that limiter or have more than 80-90% CPU or GPU load. Half of the "new" settings are literally renaming old driver settings to a non technical term that they can market... Ultra low latency = pre-rendered frames 0 aka JIT rendering, low latency on = pre-rendered frames 1 aka what everyone sensible has been using for decades, Reflex Boost = prefer maximum performance aka what everyone sensible has been using for decades. Unfortunately the only thing they added was reducing latency in GPU limited scenarios, which is something that shouldn't exist in competitive games, there should be spare CPU and GPU performance and never be limited other than by an FPS/processing limiter in software. prerendered frames 0/JIT/Ultra is good for Vsync gaming aka the few non competitive games that only run well with Vsync.

What I would like to see is a test that is done in the same software/game, not different games for each test. And include all scenarios, GPU un+limited, CPU un+limited, Vsync ON/OFF, fps limiter ingame ON/OFF, Reflex ON/OFF, but of course this blows up the number of tests and size of the data table to show. The current method makes it hard to compare and always misses some useful use cases. Currently it looks like Reflex does almost nothing better latency wise compared to an FPS limiter set to the same FPS value, plus who the hell runs no FPS limiter these days when most monitors have Async, no one should, nor should they run GPU limited, and that makes Reflex obsolete in GPU unlimited scenarios. And games like say Metro/Control/... RTX etc. that do run easily GPU limited do not support Reflex... Where as GPU light online games that do not run GPU limited support Reflex... they really got it backwards at Nvidia in terms of useful support just so they can market it easier to masses in these silly kiddie games.

Is Nvidia's Reflex, anywhere close as if you downclocked your cpu, Would you get similar results, in any way tho, how Nvidia Reflex is described, the same thing would be described if you downclocked your cpu. I don't know to what extent, but there's probably ways to find that out too

So if the game Don't support reflex use the Nvidia latency ?

numerically the latency mode does make your pc responsiveness in ms improve by margin, but afterall i stopped using it because for my system i can tell the mode itself is buggy, it makes queuing in the system wild by acting like telling the heart to skip a beat here and there. and working with vsync it makes the screen choppy, not smooth. i turned it off so i can enjoy my computer again. am a decent gamer myself the smoothness is well preferred, that 20ms improvement does not hurt my experience as much as choppiness. the low latency mode also causes some unwanted lockup due to the bugs i experienced

My question is .. Do i get input lag, stuttering if i enable reflex mode and null ?

이런 전문 팩트 결과를 내놔도 마케팅에 속아서 무조건 켜놓기만 하면 적용되서 반응속도가 빨라졌다고 말하는 자칭 전문가들이 많지요. 자칭 전문가들이 가장 많이 하는 말 "빨라진거 같다" "빠릿해진거 같다"

I think consistency is way more important than low-latency. While I agree with your methodology and you guys did a great job with Reflex Mode, you overlooked consistency of mouse movements. When I turn on Reflex Mode, I can notice a huge difference when I try to flick to a target. My flicks never go where they're supposed to. They always end up slightly too short or too far. I played for a couple of months with reflex mode on before I realized this inconsistency was happening. Once I turned Reflex Mode off, I instantly felt the difference. My flicks were way more consistent. Now if I miss, it's because I messed up or the enemy suddenly strafed the other direction, not because my cursor ended up in a different spot than it should have. This is ultimately the problem with Reflex Mode and why you shouldn't use it.

I would suggest putting the mouse on a track or something and measuring mouse movement versus character rotation. I predict that you will find a huge discrepancy when Reflex Mode is enabled. Here you would get a much more important metric than just latency. Character rotation is way more important than low-latency clicks.

Not knocking your findings so far, you did a great job, keep it up.

This is all info Nvidia should have given us. Thank you for your work!

Insane..
I remember times when removing Screen Tear was the most crazy shit ever..

But this is just next level another one on top

This feature exists for like 10 years now in nvidia(pre-rendered buffer frames), they just renamed the option and selled it like a new feature....all comercial dude

Neeeed.... Tarkov tesssttttttt.......... Please Lord Battlenonsense the breekis beg of you

your videos are perfect for someone like me who has thousand of other obligation but still would like some nice monitor or mice when i have free time to play and i do not have to do day long research on the thing, thank you from an ship navigation officer.

In NVIDIA settings so which one is better for low latency mode: Ultra or ON ?

Please test Black Desert Online with 60 FPS Cap on what's the best settings you can recommend. thank you

Could someone summarize for me? I'm Brazilian and I don't understand English

In the most recent driver revisions I see that with a G-sync/G-sync compatible display and G-sync enabled within the nvidia control panel, enabling V-Sync along with Reflex in game sets a frame rate limit below your monitors refresh rate (158fps on a 165hz g-sync display). I can only confirm this on a 30 series GPU. Do you have any confirmation that this does not add any additional input lag?

So always have it on?

Should i use nvidia reflex low latency with directx12?

Insane video, so much credit due for the work in this video! Thanks for helping me understand! Absolute legend!

is there screen tearing when no-cap fps and nvidia reflex=on ?

so what is better?
1) rtss=on at max hz-3, freesync/gsync on, vsync off
2)just nvidia relfex on
3)rtss=on at max hz-3, freesync/gsync on, vsync off and nvidia reflex on.

It would be interest what happened with reflex off and fps capped to 67 (gpu load about 98%)

The thing I'd really like to see is how this affects latency with g-sync on/off with fps limiters below, at, and above the refresh rate of the monitor.

Maybe I don't get it, but why is this a separate SDK that needs to be integrated with the engine? As I understand it, all you have to do is to limit the number of submitted render commands, making sure you keep the GPU fed close to 100%, but almost never filling the render commands buffer. Can't this just be implemented directly in the engine as a custom solution, without any nvidia sdk? The only reason I see you'd want the SDK is to interact with nvidia's analysis tools, otherwise you'd have to use a high speed camera and go through the old time consuming process of gathering & interpreting data yourself. Even so, a custom implementation would be better and could be multi-platform, whereas nvidia's sdk limits you to their platform.

Would enabling gsync make a difference to the latency if you have RLL enabled? Would gsync not require a framerate cap in such a situation?