Input Lag and Frame Rate Limiters 

Whoa. Thank you!

Wow. Thanks!

THIS

Thanks! For Quake, you may just want to find a used CRT for sale. :) I haven't been keeping up with monitors lately, but I still really like the ASUS VG279QM. For chasing that CRT-style image, though, I think something newer like ViewSonic's XG2431 might be better. Haven't personally seen it, but it has the newer Blur Busters certification, which means it can strobe at any refresh rate you'd like, even 60 Hz.

Adaptive sync is universal name of gsync and freesync.

Using RTSS like that is a pretty clever solution to use for games that don't have internal caps. If the in-game limit is set lower than RTSS, RTSS won't be doing anything, so it's safe to leave it on generally.

It's better to cap fps and turn vsync on and even low latency mode all in game engines

The way I tested lag with my super simple UE4 build can't really simulate a CPU limited situation. I don't expect it would make it any worse, though. I'm using a 7700k for my main system, but I also have an Ryzen 1600 on another PC, so I'll see if I can find some way to test being CPU bound properly!

I need to buy an AMD card at some point to test all this, but BattleNonsense found Enhanced Sync to work just as well as NVIDIA's ULLM. And Fast Sync should reduce latency if you can get a frame rate that's an integer multiple of your display's max refresh rate. I may test this for the next one. How are you running your CRT with the 5700?

@@ApertureGrille thanks for the info, I somehow missed the Fast/Enchanced sync tests from Battle(non)sense, will check them out. Right now I'm using a cheap 240 MHz 6bpc DP-VGA adapter, but I also have a 330 MHz delock 62967 ordered. There are even better adapters such as Delock 87685 and Sunix DPU3000 which are able of 400+ MHz pixel clock.

@@elesinalex Your Lacie can do 140 kHz horizontal clock, right? I'm jealous. The 330 MHz Delock should be good enough for resolutions below 1600x1200.

For games that allow it, you can set an internal limit fractionally below the strobing refresh rate. So for instance, if you're strobing at 120 Hz, with VSync, you have to eat a bunch of lag, but if you limit via the game to 119.9, you can reduce lag at the cost of a consistent hitch every few seconds. You can spread those hitches out by limiting closer and closer to the fixed framerate, e.g. to 119.999.

@@ApertureGrille Thank you kindly. Chris of Battle(non)sense not too long ago released a video regarding frame capping with g-sync for lag reduction, and recommended 4 frames below the max refresh of the monitor, or possibly more with higher refresh monitors. Is just capping the framerate to 1 frame below still good?

@@superbro6413 I tested 1 frame below and 3 frames below and found no difference. Perhaps there are some games or engines where you'd want to cap a little extra below if their limits aren't very consistent, but with something like UE4, its cap is steady and accurate. Seems like a "playing it safe" kind of rule, but there's really not that much difference between getting 164 FPS and 161 FPS.

@@ApertureGrille Thanks again.

Gets darker? That should only happen if you're using a blur reduction mode. For G-Sync monitors, NVIDIA calls it ULMB (Ultra Low Motion Blur). ULLM is NVIDIA's frame-rate limiter and lag reduction setting. Which monitor do you have? Good suggestion about another video! I wanted to look at whether using a mouse with high DPI and low sensitivity was better than the other way around as well.

​@@ApertureGrille ​​ Looking forward to the DPI / sensitivity video! From what I've seen so far, using as high a DPI setting as the mouse sensor can reliably support while lowering in-game sensitivity settings has two benefits: 1 - Reduced input lag brought about by using higher DPI. This has been tested by both Optimum Tech ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-imYBTj2RXFs.html and Battle(non)sense ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-6AoRfv9W110.html using Nvidia's LDAT. 2 - Increased granularity / precision in aiming granted by lower sensitivity. This is the one that's a lot more interesting to me, even though it's probably less impactful in actual gameplay. I haven't found a comprehensive explanation, but piecing together what I've seen from 3kliksphillip's video ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-8XNUp70mDlQ.html and these great demonstrations in CS:GO ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-NUiGkDB_48s.html ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-tsHV_ZH3T9c.html, this is what I've come up with: In 3D games that use rectilinear projection such as first person shooters the in-game sensitivity determines rotational movement by applying a certain amount of change in angle per count registered by the mouse - i.e., x degrees movement per count (or "dot") of the mouse's sensor. Since there are discrete steps in this angular movement, we can determine that there are certain areas in the 3D space that we cannot directly aim at, and that this problem becomes exacerbated as the distance increases. Low sensitivity therefore helps us by reducing the angle change per count, making more of the 3D space accessible to our aim. In turn, and in addition to the latency benefit, raising the DPI allows us to lower the sensitivity while maintaining the same relationship between physical mouse movement and the corresponding change in game. In an ideal world, we would maximize the latency benefit by using an infinitely high DPI and lowering our in-game sensitivity to get to our preferred feel. On the other hand, to get the most precise possible aim, we would use an infinitely low in-game sensitivity value and raise mouse DPI correspondingly. However, neither are really practical since mouse sensors have a ceiling after which they begin to have problems with accuracy (or simply cheat by multiplying the registered counts given to the computer). So practically speaking the best course of action is to pick the highest reliable DPI setting (1600 or 3200dpi for most sensors) and adjust in-game sensitivity to accommodate. Of course, the benefits of high DPI / low sens are mostly academic since with any reasonable sensitivity the inaccessible space would be way smaller than any hitbox anyway - and empirically most pros are doing perfectly well with the "high" sensitivities used with 400dpi, but it appeals to the nerd in me regardless. I could also be terribly wrong about all of this, so knowing how meticulous your are with your work I'm really looking forward to seeing your take :)

I don’t think so, he said in the last minutes to turn on ULLM and frame cap.

Reflex is awesome at removing much of the latency penalty when you hit 100% GPU use, but it's still worse than not running into that GPU limit at all. If you want to use both at the same time, you may want to use an internal cap to something a little lower like 220 FPS. That's only a render time difference of ~0.4 ms, but it should keep you away from maxing out the GPU. Hopefully that'll keep you below 90% GPU use, but Reflex will be able to take care of the situations when your GPU use hits 100%.

No. You'll get the benefit of reduced latency even on a 60 Hz monitor if you let the game render as fast as it possibly can!

im running a older 144hz screen that doesnt support adaptive hz and im runnig csgo with around 300-500 fps would it be better to cap it to 144fps?

@@BlackgoldTwinduck In that case, no. If you're OK with the tearing, the more FPS the better! I mentioned in the video that I play Quake 2 at 1000 FPS, V-Sync OFF. With that, at least some portion of the screen is showing my latest inputs every 1 millisecond. Take a look at 2:30 in the video to see the tearing.

Got something different coming soon, but I'll revisit input lag and Reflex some time after.

@@ApertureGrille nice, can’t wait!

Yes! The drop in latency there has fully to do with the GPU render time with different rendering resolutions. For example: when limited to 60 FPS with a GPU use of only 56%, the GPU render time (how long it takes the GPU to render the frame) is at around 9.2 ms. When I drop the rendering resolution down, the GPU has an easier time rendering the frame, so at r.ScreenPercentage of 20, the GPU frametime is only 2.2 milliseconds, which is why the GPU use drops to 7%, but that GPU render time difference is where the latency reduction is coming from. At lower resolutions (or if you had a mega-powerful GPU), the GPU can render the frame faster, so the final rendered image can show up on the screen with less latency. This is really the same as measuring latency at a higher framerate. Higher is always better.

VSync was enabled for all the tests, otherwise, during VRR, you'll see screen tearing at the bottom of the screen. Enabling V-sync doesn't increase lag in the VRR range, it just removes the tearing.

@@ApertureGrille Ahh ok thanks for answer!

@@ApertureGrille Please add this to the description since it was not fully clear in the video. Thankfully I found this comment :)

Thanks! I'm generally skeptical when people claim certain settings introduce lag. There are, of course, the big three: V-Sync, 100% GPU utilization, and framerate. These actually have a large impact on lag, compared to something like anisotropic filtering, which should have almost no impact. The fps_max setting might make a difference, but... I doubt it. :) I can look into it, though.

@@ApertureGrille Thanks for the reply and thanks for your time!

@@ApertureGrille Hi again, have you had any success checking out the fps-setting?

I'm sorry if spam this section :D

@@ApertureGrille Hope everything is fine! And that you don't think im to persistent! :D

for strobing modes capping 59hz will give you stuttering but 59.99 does not.

@@brett20000000009 That sounds like it would work, but some engines allow decimal limits. For UE4, you can do t.maxfps 59.99 and it'll cap to that.

@@ApertureGrille Thanks for the reply. I have been playing cyberpunk 2077 it's framelimiter won't do fractional values. I have a single strobe monitor and the vsync lag at 60hz is horrible. I just tried 60.01 with a 60fps ingame cap and it feels way better. :) feels better then rtss.

Adaptive-sync is amazing for most games, but if your graphics card doesn't support it, I'd suggest turning V-Sync OFF and letting games run as fast as you can. I've got a video coming soon talking about V-Sync OFF, but in summary: for the lowest lag, turn V-Sync OFF, run your games at the highest framerate possible that doesn't max out your GPU.

@@ApertureGrille Graphic card support adapted v sync not the free sync. MY monitor is free sync.

@@ApertureGrille By adapted sync u mean adapted v sync or I got everything wrong?

upload more videos with more precise information please. :(

@@Mobiledash52 Adaptive-sync, G-Sync, FreeSync, Variable Refresh Rate (VRR), are all terms for the same thing, but you need at least a Pascal (NVIDIA GTX 1000 series) graphics card to enable it. It won't work on a GTX 980 Ti. I don't quite understand your setup. What is your graphics card and monitor? If you don't have a Pascal (or newer) graphics card, VRR/adaptive-sync/freesync/g-sync won't matter. Just turn V-Sync off and run at the highest framerate you can.

In that case, I'd turn it off. ULLM's cap is pretty aggressive, so it would override the internal cap, and internal is better for latency. For instance, at 165 Hz in UE4, ULLM is capping my framerate to 157, so an internal cap of 164 or 163 wouldn't be doing anything. For games that have good internal caps, external caps are unnecessary and should be disabled. I know I said in the video that you might just leave it enabled globally, but lots of games don't have internal limiters.

@@ApertureGrille Ah, thanks for the response and info.

@@ApertureGrille Just to confirm though, ULLM is a better alternative to simple external FPS limiters (such as NVCP or RTSS) if there is no internal frame-rate limiter in the game, such as War Thunder? Or should I still just use NVCP to cap FPS instead of ULLM?

Thank you! Adaptive-sync, variable refresh rate (VRR), FreeSync, and G-SYNC all are different names for the same adaptive monitor refresh rate behavior. Adaptive-sync has been one of the most impressive tech improvements for monitors in years, and I'd always recommend gamers use it... unless you're going to the lowest input lag possible. I mentioned in the video that I play Quake 2 at 1000 FPS V-Sync Off because that's one of the most immediate and responsive games around; that generally holds, and I've got a video coming out about V-Sync off soon. If you're serious about reducing latency, you should do three things: 1) run games at 300+ FPS, 2) turn V-Sync off, and 3) never let your GPU go above 90% utilization. For most people, though, losing a few milliseconds of latency isn't worth getting all the tearing and stuttering that comes from running V-Sync off.

​@@ApertureGrille Thank you for this precious information, I mainly play Valorant at the moment. Based on your video I did activate GSYNC and set my in-game FPS limiter to 357 FPS since my monitor refresh rate is 360hz. I also turned off VSYNC in-game and in NVIDIA parameters. I guess this is already a good setting? Now if I want to go low FPS, uncapped the FPS in-game, and keep GSYNC? By the way, my FPS goes between 450 and 750 it's not consistent, does that matter? Also, I don't really see any screen tears or stuttering. Looking forward to your next video! Once again, very well done!

@@boa974 if you don't notice the tearing, you can let your framerate run as high as possible. Normally I wouldn't want to let the framerate swing that much, but 450 vs 750 FPS is only a frametime difference of 0.9 ms.

@@ApertureGrille Could you test the difference in 60hz and NTSC 59.94Hz if it affects pacing? and for VRR if one has to limit at 58? maybe 119.88 | 120Hz By intuition X/1001 modes will free a little GPU utilization/time!

Tecware Phantom RGB.

If the game has an internal limiter, use that. But if not, ULLM does as well as any other external limiter but also helps in situations with 100% GPU utilization.

a5hun do i need to use NV drivers fps limiter with ULLM? Monitor is vg27, bought in with big influence of ur review, thx ;)

@@barabulkafication Check 3:53. ULLM will work in DX9 and DX11.

@@ApertureGrille kk, thx for answers!

ULLM does, but LLM does not. I didn't find any case where LLM had any benefit.

@@ApertureGrille so i put ultra low latency mode globally? thanks! nice vids! hope you make a video about ping , lag etc.

Haha. Thanks! How did you find the video, btw?

@@ApertureGrille I think I was just wandering in RU-vid for input lag related videos. Tried to find a new way to fix input lag. All the internet yells, that RTSS is the best fps limiter, but it turns out, that it's not that simple.

If you don't have a VRR/high refresh rate display, you're pretty much in the same situation as us CRT users. The best way to minimize lag is to turn V-Sync off and live with the screen tearing. But visually, that looks really bad; a 60 FPS game running V-Sync Off 60 Hz will have tear lines that linger and bounce up and down around one portion of the screen. It looks awful. But because there is no render queue, latency is the lowest you can get for 60 FPS. Next best would be to use RTSS's scanline-sync. With that, you leave V-Sync off, but RTSS tries to throttle the CPU so that it presents a frame just in time to be rendered by the GPU. Scanline-sync works really well, as long as your GPU isn't under too much load. You could also enable V-Sync and use the in-game cap to limit to 59.99 FPS. If you limit to just 59 FPS, you reduce the backpressure from the render queue, but you'll get a stutter from the missing frame every second. Limiting to 59.9 has the same lag benefit, but it will only stutter every 10 seconds. I haven't tested how many decimals you can cap just below 60 to keep lag down, but you may not even notice the stutter at 59.9. But really, the best way to limit lag at 60 FPS is to get a high refresh VRR monitor :)

​@@ApertureGrille Dumb question: I imagine it's possible to have a VRR monitor run at a high refresh (eg: 144hz) while playing a game with a max framelimit of 60fps, yes? Since it appears that running a game at a few frames lower then a monitor's max can cut down on lag, then does that mean that any high refresh VRR monitor could play 60 fps games at with lower latency then 60hz monitors? Thanks again for all your help, appreciate the informative videos!

@@superbro6413 Yes! ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-dPMHEyz38TM.html That chart may be a bit unrealistic because a simple UE4 build isn't really doing much work, but the results at 60 FPS on a 165 Hz VRR display are incredibly good. Two things work in the VRR display's favor: 1) the GPU no longer has to wait on the display, so you get rid of V-Sync backpressure while also getting rid of the annoyance of having to use internal or external limiters, and 2) high refresh rate displays scan out their panel from top to bottom faster. For a 60 Hz display, the bottom of the screen is always 16.7 ms behind the top of the screen. For VRR displays, a 165 Hz monitor will scan out the screen from top to bottom in only 6.06 ms (even when displaying 60 FPS content!), so more of the frame gets to you faster.

@@ApertureGrille Thanks again! All this information is very helpful.

FreeSync is AMD trademark, "VESA Adaptive-Sync" is the "generic" name for FreeSync on HDMI and DisplayPort, and of course that is not to be confused with Adaptive V-Sync.

I'll look into it. I'd be skeptical of any type of reflex-meter, though.

@@ApertureGrille I'm completely skeptical about it, even though I only looked at the "best result" to have something to compare (and there is no way I improved magically my reflexes by 20 ms between two tests, I'm regularly doing them, it's part of my "fps" routine to check my fatigue level). Anyway that's why I'm requesting help. I sent you an email too, so feel free to ask anything else.

@@JonasHopp what was ur solution btw

@@ShoxSerious Without any proof or real solution, I liked more the perfect consistency of capping with RTSS. But, and I don't know why, I was still seeing tearing artifacts if I capped at 279 hz. under 240fps/hz it's perfect though. So even the screen can go to 279 with VRR, I capped at 240hz in RTSS. Test for yourself if you prefer fps_max or RTSS capping.

@@JonasHopp i was trying to use the fps max, but when i use this command my frame time varies a lot, tks by ur answer

V-Sync off, from the perspective of the swap-chain, is essentially lagless. As soon as a frame is done rendering, the front and back buffers are swapped, and you'll see part of that frame on the screen... with a tear line. That means the lag you "feel" depends ONLY on the game's render time. For instance, I usually play Quake 2 at 1000 FPS V-Sync Off. There are multiple tears on every frame, but a 1 ms render time makes the game feel very snappy. From there, you'll just have a linear increase in "lag" based on the framerate. 500 FPS gives 2 ms of lag, 250 FPS gives 4 ms, etc. So for V-Sync off, higher framerates are always better, but there's not much for me to test. That's why I said in the video that V-Sync off isn't all that interesting :) Try grabbing q2pro or ezQuake and playing at 1000 FPS. It's really good!

@@ApertureGrille thanks for the clarification

Without knowing your setup and game, it's hard to say. I've seen similar things when a second monitor is connected while using VSYNC. If you are running two monitors, try disconnecting the secondary.

a5hun I have only one monitor i play on laptop

i use my laptop monitor