Input lag and Gsync on laptops

Oh look, I'm posting a thread. Yes, amazing, I know. I don't even have a problem that needs fixing this time! Anyway, onto the meat of the matter.

So this dude that does a lot of hard data tests called Battle(non)sense made a video testing input lag on mobile IPS panels and with gsync. Was pretty interesting, the tests. So without further ado, here you go!



Of importance to note is how to actually utilize gsync properly, and the fact that vsync appears to be on in the driver by default. Especially important to @Phoenix who can't live without Gsync.

Also, I figure some of you other peoples may find good use for it.

@Johnksss@iBUYPOWER may find this data interesting, though it does not directly affect overclocking.

Enjoy peoples

 

I was always under the impression that you should leave vsync on in the NVIDIA control panel and off in whatever game you are playing in order to have G-Sync working without needing to worry about limiting the frame rate. This is interesting new information, and something I'll want to check out for sure.

 

I'm going to have to check that out further, because when I've fussed with G-sync in the past, unless something has changed, if you set G-sync ON and set v-sync off and turn on the G-sync indicator, the indicator does not show. Plus it does not improve visuals when using the G-sync pendulum demo which is readily apparent when G-sync is on or off.

The other problem is if you have to limit framerate, Nvidia doesn't readily offer this option, and most games don't include it. There is Nvidia Inspector which will allow it, but why need another program for this, which is ridiculous IMHO. It should be part of those pull-down menus in the drivers.

If this is the case, then I think we need to put in a request at Nvidia forums to include a frame rate limiter in the driver options.

 

That's a good idea. They really need to offer such an option in their package regardless if it can be done through another program.

 

It's not like the driver does not have this feature already; nVidia Profile Inspector only adjusts options in the driver. If the driver does not have it, it cannot be used, so... there really is no excuse.

I will say though that the driver limiter, strictly speaking, adds more latency than using RTSS or an in-game cap. It is not nearly as bad as vsync but he tested it in an earlier video.

Sent from my OnePlus 1 using a coconut

 

How can RTSS be different than driver limiter? It's using the same stack. In-game cap I can almost see because it's implicit to the game, but RTSS can't be using anything different than Nvidia Inspector.

 

There must be something different, because the raw numbers say they do.

 

so what's the consensus now D2 Ultima?

Shall I disable VSYNC otherwise it would take over and GSYNC won't work?

sent from my coconut

 

Basically. In-game limit your fps or use RTSS if not (I still find that NVPI limit is doable but if you want the best performance for action games then you know what to do) and you're good to go. 118 fps is a good lock for gsync, most likely. 116fps if not.

Sent from my OnePlus 1 using a coconut

 

This guy does great work and I love his videos however there is a case where fastsync offers an advantage.

Fastsync and vsync off are the only rendering modes where if your fps is higher than your monitors max refresh rate then the gpu will NOT wait for the screen to refresh to start work on the next frame.

When you use vsync or gsync and your framerate is higher than your monitors max refresh, the gpu has to wait. And while it is waiting, the cpu can be waiting for it to accept more draw calls.

Using the in game limiter can be a good option but in some cases that will give up performance that could lead to lower input lag.

The 1070 can render a frame in about 2ms average in competitive play in overwatch on the lowest settings. At 144hz the monitor takes almost 7ms to refresh. So when the gpu finishes early, it starts waiting. Then the cpu accepts input from the mouse and as soon as it trys to give a draw call to the gpu, it starts waiting as well. So the input for the next frame ends up being delayed even more and this can be compounded for every technology such as overwatches buffering option, tripple buffering, and even multithreaded game code in general.

Mind you that most of the people who would use low settings on high end hardware are pro gamers or people who are just really competitive. But in these cases you need an unconstrained rendering mode to leverage the advantages of the higher framerate and your only choices are vsync off and fastsync.

The stuttering is a result of the fact that that: The delay between the last frame rendered and the last refresh is not regulated by the gpu waiting for the screen to refresh ever. So the delay between animations and display can vary by as much as one frame. In cases where the framerate is close to the monitors refresh rate, then that is not acceptable stutter. But when you're talking about a 3.33ms total frame time (300fps), that stutter is less noticable.

It all comes down to the fact that not many people (even the guy making this video) consider the case of extremely high fps when testing this technology even though it is intended for use in that context.

 

Why don't you mention this to him and ask him to check? Though remember as far as I know, fast sync only works with single GPU. If you need high fps in a game and one card isn't cutting it (idk maybe you like the eye candy like me) then fast sync is not an option

Sent from my OnePlus 1 using a coconut

 

Play battlefield 1 on low, and then play it on max. Tell me which looks better

 

NVIDIA needs to implement something like Radeon Chill to give users access to a built-in frame limiter, with the addition of universal compatibility, of course.

 

You can force fast sync with nvidia inspector even for sli but i believe there is already a stutter introduced by sli. For the reasons you listed as well, i dont think that anyone would want to use fast sync with sli anyway.

A test comparing fast sync with 300fps and low graphics settings, to vsync off, to frame limited gsync, to 300 fps with fast sync and vsync would be great.

If you can get the fps high enough, fast sync would be better but a real world test would be really helpful to know if that can even happen in overwatch. I know the technology was designed with csgo in mind so i believe in that game it can help with ultra high fps.

 

When he made a set of videos abouy Vsync awhile back I tested many settings and ended up tunring Vsync off and used RTTS to limit my frames to the refresh rates, 120 on mt GT73 and 144 on my desktop, they have been pretty good but today I tried 139 on my desktop and I notice that BF1 was much smoother, so I'm sold on Gsync on Vsync off. It works.

 

I tested Crysis and Call of Duty: Infinite Warfare, both at 4K resolution. By limiting the frame rates of both games with RTSS and having vsync off in game (leaving the control panel to the 'let application decide' default), there was a perceptible decrease in input lag versus vsync on in control panel and off in-game. The guy who made the video is on to something.

 

I'll have to check that out. And perceptible difference could also be placebo effect. I don't doubt it may be effective, but I'd really like to understand and know why RTSS is different than something like Nvidia Inspector. How do they limit fps differently? I honestly don't like RTSS because it has given me fits in the past, and it has to be active while running frame limiter where Nvidia Inspector does not. I can set it to an FPS universally and it's set indefinitely whether Nvidia Inspector is running or not.

 

NVPI limiter is a driver-based limiter, not related to any sort of hooking.

RTSS (or OBS, which has a similar functionality) is a hook-related adjustment, and should be different in nature, just like how limiting FPS in-game is different to the driver.

What I don't get is why the FPS limiter in the driver is so laggy. I mean, I have found extreme usefulness for it, certainly, but if there's going to be an eventually-notice-able lag scenario it makes sense. Come to think of it, this may be possibly why when I had to use my second monitor only and I played Overwatch in Borderless windowed and limited FPS via NVPI the lag felt beyond unbearable. I know the monitor had notice-able input lag on its own, but OW felt exceptionally bad. It's interesting food for though.

I checked his borderless window and windowed mode scenario and the input lag with windowed/borderless and vsync is up in the vicinity of over 120ms for 60Hz. Apparently 120Hz+ is a different story, and it's probably in the bearable state, but I think I'll default to using in-game FPS limiters where possible instead of relying on NVPI.

RTSS though... I'll have to think about it. I would try to bring up the issue to nVidia and ask if they could improve the FPS limiters, but... well... we know how much they care about us

 

If I were running on a high-refresh display, that would be a possibility. However, I was testing at 4K/60Hz, and during times where the frame rate would drop below 50, there was a difference between having the limiter on and off.

 

Yeah, it seems the driver limiter should be the best implementation. Maybe we should open a thread at Nvidia forums and we all pipe in there and maybe it might draw some attention. It is kind of ridiculous if you think about it. They offer G-sync but no obvious way for "Joe Gamer" to use it, or haven't been using it to its best potential. I obviously haven't. I'd like to think it's an excuse for my bad game playing though. But I know that's not the case.

I believe you. I'm just trying to wrap my head around why it should be such a big difference.

 

If we do it, we need to not say we're laptop users. We'll instantly get attacked for it and people will write off anything we say.

If anything, we should bring up the fact that vsync is automatically globally on for single GPU, Gsync laptops (it isn't for me, but I am non-gsync, SLI) after the initial statement.

Either way, I'm down for joining it all.

 

well from what you guys are saying, i should limit my fps to 60 for g sync to work properly but not alll games got that option in their settings,. and if active V sync that will just make the perfomance worse.. i'm unsure about what to do now... confused :/

 

No, you need to limit your frame rate to something less than 60 in order for G-Sync to work. For games that don't give you this option, you can use the program RivaTuner Statistics Server.

 

Yeah, from what I've discovered using NVPI and in-game limiters, the driver limiting function picks the best frame to limit to. 62fps for 59/60/61/62fps, etc. 125fps for 119/120/121/122/123/124fps. It seems that certain FPS limits don't work very well and are very stuttery.

For example I limited Overwatch to 120fps. It looked horrible. Sometimes there were massive lagspikes.
Then I tried 119fps. Same thing. 118fps also same thing.

116fps was fine. 125fps, 126fps were fine. No stutter present.

So it seems you have to fiddle a little to find a non-stuttery framerate limiting. I think the driver limiting in NVPI is a good dry run scenario. Try driver-limiting near the fps you want, then pick an FPS limit at where the driver limiter defaults to. Since 119 through 124 limits to ~125fps, 125fps is a good limit. 132fps is also a good limit. 56fps is a good limit. 70fps is a good limit.

Fiddle and see what's best, then use it for gsync (or for driver limiting in general).

 

So people have been paying for gsync hardware all this time and it's not even working as intended?

Criminal.

 

i loved this thing i discovered.. now screen tearing is actually gone!

 

Working as intended, just not properly utilized. And it comes down to response time and if you notice it or not in the games you play.

I still put the onus on Nvidia though for not making it clear how users should best utilize it or offer a "one button optimization" method for those that don't want to fuss with the details.

 

I don't know about you guys, but my Gsync does not work without Vsync being engaged in Nvidia Control Panel. I've capped FPS even as low as 50 and still saw tearing in the bottom part of the screen.

Maybe laptop Gsync work differently from the desktop one.

 

A desktop G-Sync monitor has a hardware module connected to the DisplayPort that communicates with the GPU in order to get it working. AFAIK, laptop screens contain no such hardware and thus the adaptive refresh needs to be implemented differently.

 

Yes, I'm aware of the module. Laptop are connected directly to the motherboard therefore implementation of the tech is different.

However, it is puzzling why wouldn't Nvidia make turning Gsync on consistent between laptops and desktops.