Seeking Triple Monitor Advice

[Flamehead]

Hello all, I’m wondering if I might be able to get some confirmation or perhaps advice. I’m ready to make the jump to triple monitors and have been doing a lot of research. I have a 1080ti and would prefer not to turn down any eye candy, but am willing to turn down some settings if need be. I won’t be going SLI.

From what I understand, a single 1080ti will struggle with triple 1440p monitors, so I’m guessing this will bring me down to 1080p. If going 1080p, then I’m guessing 24” monitors are preferable over, say, 27” because of a possible “screen-door effect” when placing the monitors up close.

Is the above correct?

If going 1080p, I was thinking about maybe getting them at 144hz, but I understand even that scenario may make the 1080ti struggle.

I don’t mind going with 1080p with 24” at 60hz. I just don’t want to “settle” on that either if my card can handle more. Thoughts?

Also, I’m not interested in VR.

 

[Jens Roos]

Edit -> the part below is actually not completely correct when it comes to screen tearing, using no v-sync in-game, global nsync on and fixed fps lower than the monitor Hz works well. But I leave this part as it was anyway

Sure, no problem. I should point out that this could be much related to SLI as these most of times needs another setup than an single GPU. Without any sync active, more over tearing, there will be issues with both low gpu load micro stutters and high fps micro stutter, differs a bit between different games.

Nvidia control panel, global setting to have gsync on

SLI/Surround

Then it depends on the game a bit, but rF2, set to video and 100Hz

Nvidia inspector has some SLI fixes for each game, but also:

In game, rF2 now runs at 100fps, and if the fps drops an reasonable amount it's still fluid

For some games it seem better to use no sync in game, and force it on instead, this I had for rF2 until recently (with None above), had to change it after either rF2 or nvidia driver update (don't remember).

To be sure I actually remember this correct, I change above "Sync" from "Video" to "None" and tried rF2 and the tearing was at once noticeable, not much but still there, my fps was set to 101fps. To short to test for any stutters issues, but that is also not an constant issue.

If using sync as above, but at 120Hz and then cap my fps to 100 in game, then that will lead to micro stutters.

According to link below, gsync can be global enable (as I have) or per game, using nvida inspector setting sync tear control adaptive. To be sure here, I did a new test with adaptive selected, same result with tearing.
https://www.geforce.com/hardware/technology/adaptive-vsync/technology

So using "vsync" on a global enable gsync setup, either in game or by inspector, this will use the gsync technology. If you don't enable any sync in-game, no syncing is done and you get tearing, and maybe different type of stutter issues.

Could be an SLI and triples combo feature, that I have no idea about.

Edit, went back and tried in-game "None" and Forced-On using the inspector, that seems to work well/similar to what I have now, so could have been something else that time.

 

[RasmusP]

You got a lot right but some things not perfectly straight.
I created a video showing the important bits while using the gsync pendulum demo.
A written overview:
- Gsync will only really work when the fps are within it's range. Which mostly is 30 fps up to the Monitor's Hz.

- When FPS exceed Monitor's Hz, gsync indicator will show that it's on but it's not active!

- When vsync is used without gsync, fps will get limited to the Monitor's Hz. If fps are below the Hz, stutter but no tearing will be visible!

- When vsync is used combined with gsync, fps will get limited to the Monitor's Hz. However if fps are below the Hz, gsync will become active and no tearing and no stutter will be visible!

- When vsync is used but set to "adaptive", either gsync will overrule the adaptive vsync or if gsync is deactivated, vsync will become "no sync".

- forced vsync or application vsync are the same for these cases. Sometimes one will be smoother though, but it's independent from gsync!

Now the crucial part about input lag:
When gsync and vsync are used, like it's recommended basically everywhere, vsync won't be active up until the point where the fps hit the Monitor's maximum Hz (which can be set in windows or within applications like rFactor 2. Windows setting becomes overruled when the application is run in fullscreen mode. When borderless or windowed, the Windows setting will overrule the application!)

So when gsync is used, which can be tested by enabling the "Hz counter" in the Monitor's OSD, like in my video (often called fps counter but you can see in my video that it's a Hz counter!).
When the number is synced to the fps number (below the maximum Hz setting), gsync is 100% active. If the number stays the same no matter the fps, it's not active! Even if the gsync indicator says otherwise. Like seen in the video.

Now when using vsync+gsync and letting the fps freely run up the maximum Hz setting, you'll see the gsync indicator active but in fact VSYNC will be active, not gsync!
When the fps drop, you'll see the Hz counter dropping too. Gsync will be active.

Now to the facts about input lag:
You'll always have input lag of about one frame, since your input can only be displayed for the next frame, not the current. On top of this there is hardware input lag from your mouse, motherboard, USB etc and the Monitor's controller.
The lowest will be without any syncing.

Vsync, Gsync, fps limiter, capturing software whatever input lag will be on top of this "basic input lag".

Vsync input lag: depending on the setting of "pre-rendered frames", which at default is 3 but can be either set in-game or in the driver to 1, causes input lag according to the amount of pre-rendered frames. (It can't be set to 0, since one of the buffers would become disabled and break the normal vsync.)

At 60 fps this input lag will be:
1 frame -> 1000ms/60 frames = 16.6ms from frame to frame.
3 frames -> 16.6ms * 3 frames = ca. 50ms

At 100 fps this will change to 10ms and 30ms!

Now gsync input lag tests:
Source video of the screenshots: LINK

Green: shortest, Blue: average, Red: longest

You see that at 400 fps, no sync is active either way, since it's beyond the Monitor's Hz.
Vsync will limit the fps to 144 and the nvidia inspector limiter is really bad.

Gsync + nvidia limiter stays really bad due to the limiter.
Gsync + CS:GO internal limiter shows the real thing.
Gsync + Vsync without a limiter shows that vsync only is active!

Now to the numbers to actually compare:
No Sync + CS:GO limiter: 30-45 ms
Gsync + CS:GO limiter: 28-46 ms
Vsync + whatever: 50-62 ms

Gsync or no sync is basically identical. Tiny differences are measurement tolerances! Vsync however causes huge input lag in comparison!
Everything with the nvidia limiter can be ignored, it's crap.

Another Screenshot, different video, same guy:

Summary:
Gsync will only be active as long as your fps are below your Hz-setting and the Hz-Counter of your Monitor shows that the actual Hz are below the Hz-setting too! It will cause almost no input lag at all.
Vsync on the other hand causes significant input lag depending on the pre-rendered frames. It will always add 1 frame-lag on top though at minimum!

Now another video, self recorded with my G27 in Assetto Corsa. FPS are limited with RTSS.
Left: 100 fps @ 100 Hz with Vsync (no limit)
Right: 95 fps @ 95 Hz with Gsync (limit to 95 fps)

I inverted colours to make the wheels more visible.

 

[RasmusP]

In game, rF2 now runs at 100fps, and if the fps drops an reasonable amount it's still fluid

Additional to my technical-speech post above:
If it runs well and you like it, don't change anything. If you don't feel/see the input lag from vsync then it's all fine. I'm mostly playing at 60 fps due to my old CPU and to make the change between games comfortable.
At 60 fps vsync will cause a much higher input lag than at 100 fps.

I do not mean to sound big headed or "from above" in any way. Just tried to keep it technical and informative I'm interested in this stuff and have to do a lot of testing with logs and records at uni (electronics, oscilloscopes etc) so I'm having fun doing the same for PC settings

If I'm wrong with anything, please correct me!

Have a nice Sunday

 

[Jens Roos]

Thank you RasmusP for your post, well done. I have seen most of them before, but informative about these things.

I think I have understand how it should work, but I can only notice that if vsync is turn off, and gsync is on (global or by inspector), the screen tearing is obvious. It's not much but still there, have not tried with gsync off.

When I was figuring out best working settings, there seems to be limitations/difference when working with SLI so it might well be related to that. For heating issues, I have side monitors in 1 card, and center in the top card. Could be an reason as well, even as nvida claims you can put them anywhere. As I will mount watercooling on my gpu:s very soon (have all the hw) I can put all monitors in the top card, I try some when it's done and see if anything changes.


Regarding input lag, IMO it's more critical in FPS games as they have been testing. I only race triple pancake and never uses an virtual wheel, and at 100 Hz everything is running smooth. I have never felt that any issues of the input lag. But both tearing and stutters are issues i'm get very irritated about on other hand...

The aim of my post was to show expected performance with 7680x1440 triples, but it's took quite a while before it was actually working, lets say I was at least a bit disappointed after investing in (back then) high end gpu:s/monitors and the performance was basically crap. But now it's working very good.

 

[RasmusP]

I think I have understand how it should work, but I can only notice that if vsync is turn off, and gsync is on (global or by inspector), the screen tearing is obvious. It's not much but still there, have not tried with gsync off.

When doing this, do you limit your fps? I saw bad gsync implementations that stuttered but I never saw tearing... Sounds like when you disable vsync, your fps will go higher than 100. You should limit them at 97 (limiters mostly have some fluctuations and if a fluctuation goes over 100 fps, gsync will deactivate).

If you limit the fps: which tools did you use? Not sure if rfactor 2 has an inbuilt limiter and the one from the nvidia inspector is crap.

 

[Jens Roos]

When doing this, do you limit your fps?

I have tried both using the profile inspector and using rF2 limiter, when I tested now it was with rF2 limiter. 120Hz is the surround limit, but I did have the rF2 monitor setting at 100Hz and limiter at 101.

I make a new test changing the rF2 monitor to 120Hz and keep the 101 fps and see if it changes things.

 

[RasmusP]

I have tried both using the profile inspector and using rF2 limiter, when I tested now it was with rF2 limiter. 120Hz is the surround limit, but I did have the rF2 monitor setting at 100Hz and limiter at 101.

I make a new test changing the rF2 monitor to 120Hz and keep the 101 fps and see if it changes things.

Changing Hz might have side effects with sli, monitor OC etc etc. Better put the limiter to 95 and see how it goes while keeping everything else like it is now!

BTW, I can't find any fps limiter within rfactor 2. Where do you find it?

 

[Jens Roos]

I have now tested by changing my monitor Hz in-game to 120Hz, gsync global on, disable the in-game vsync and set the in-game fps to 100Hz. And you are absolute right here, it works very well in AMS, rF2 and AC I have put an little edit in my previous post, but kept them as they where written

I still have to test if low gpu load stutters still there, but it was mostly rF2 and there it has happen a lot the last 6 months with the gfx engine as well.

You find the fps limiter in the player.json file:

Edit, after testing a bit more, rF2 suffers from low gpu load stutter with above settings, one solution is setting the refresh rate to 100Hz and "sync" to "video" as I described before. But this maybe not the main thread topic more than, don't run high res triples with sli unless you like to fidle a bit

 

[Flamehead]

Hey guys, this is fascinating discussion. RasmusP, from what I gather it sounds like you're saying the following:

In a 144hz monitor scenario, gsync won't be in effect when the fps exceed 144hz, so there WILL be tearing above the refresh rate. One way to remedy that is vsync, but that's not preferable due to lag. Therefore, the way to go is to set a frame limit (using something like RTSS) to, say, 144hz--or maybe a hair lower.

That way, frames never exceed the monitor's refresh rate (resulting in no tearing above 144hz), and with gsync, there is no tearing or stuttering below the refresh rate (at least down to the minimum gsync range of 30hz).

Is that about right?

 

[RasmusP]

I have now tested by changing my monitor Hz in-game to 120Hz, gsync global on, disable the in-game vsync and set the in-game fps to 100Hz. And you are absolute right here, it works very well in AMS, rF2 and AC I have put an little edit in my previous post, but kept them as they where written

I still have to test if low gpu load stutters still there, but it was mostly rF2 and there it has happen a lot the last 6 months with the gfx engine as well.

You find the fps limiter in the player.json file:

View attachment 286205

Edit, after testing a bit more, rF2 suffers from low gpu load stutter with above settings, one solution is setting the refresh rate to 100Hz and "sync" to "video" as I described before. But this maybe not the main thread topic more than, don't run high res triples with sli unless you like to fidle a bit

Is your power management setting set to "prefer maximum performance"? I had some massive stuttering at low gpu loads with the default "adaptive" which got updated to "optimum" 2 years ago.
Check your gpu clock speeds. My 1070 dropped into 2d mode with rF2 and AMS, resulting in stutter although in theory, it would've been fine.
Nice that it's working though!

 

[RasmusP]

Hey guys, this is fascinating discussion. RasmusP, from what I gather it sounds like you're saying the following:

In a 144hz monitor scenario, gsync won't be in effect when the fps exceed 144hz, so there WILL be tearing above the refresh rate. One way to remedy that is vsync, but that's not preferable due to lag. Therefore, the way to go is to set a frame limit (using something like RTSS) to, say, 144hz--or maybe a hair lower.

That way, frames never exceed the monitor's refresh rate (resulting in no tearing above 144hz), and with gsync, there is no tearing or stuttering below the refresh rate (at least down to the minimum gsync range of 30hz).

Is that about right?

Yes, that's correct! Blur busters recommend 3 fps below refresh rate. You really don't want the fluctuations from the limiter (even rtss has some hiccups sometimes) to shut off gsync.
Having tearing for 1-2 frames wouldn't be a big problem but gsync needs a little moment to kick in again

 

[Flamehead]

Excellent, thanks! I've learned a TON from this thread!

 

[Beef36]

.

 

[BenKay]

I've discovered in RF2, that if running windowed or borderless, with G-Sync on (and set to work fullscreen or windowed), then G-Sync is actually not activating with RF2 (as seen on the monitor OSD, no fluctuating HZ rate).

I have to run RF2 in full-screen mode, then G-Sync kicks in every time....the "G-Sync indicator" you can turn on in Nvidia Control Panel, was telling me G-Sync was active even in windowed or borderless mode, but this clearly was not the case.

I don't know if this is some kind of bug or not, but it shows how crucial having the monitors' own HZ rate OSD turned on is.

 

[RasmusP]

I've discovered in RF2, that if running windowed or borderless, with G-Sync on (and set to work fullscreen or windowed), then G-Sync is actually not activating with RF2 (as seen on the monitor OSD, no fluctuating HZ rate).

I have to run RF2 in full-screen mode, then G-Sync kicks in every time....the "G-Sync indicator" you can turn on in Nvidia Control Panel, was telling me G-Sync was active even in windowed or borderless mode, but this clearly was not the case.

I don't know if this is some kind of bug or not, but it shows how crucial having the monitors' own HZ rate OSD turned on is.

Yes, when I'm configuring a game I always activate the Hz-counter for a while for exactly this reason! The indicator only tells you if the GPU activated gsync but it can't tell you if the Monitor is actually reacting to it.

I have to say I play all the games I run with gsync (which are basically all, lol) in fullscreen. The windowed support was patched in and honestly, it's broken for me in 9/10 games. Stuttering, scanlines, flickering...
rFactor 2 is one of the few games that actually run flawlessly in borderless with gsync!

Alt+tab out of a game in fullscreen sure takes a little longer but it's not that bad. Quite some games jump into borderless or windowed mode when you tab out of them anyway.
So if you want to read something or whatever while waiting for a game-match to be found or between sessions, you'll still see what's going on.
rf2 and AC jump into Borderless for me, Dirt Rally and Battlefield switch into a little window.
Rocket League though minimizes for real but the icon will flash when a game is found

I guess if you're used to borderless it will feel like an eternity to alt+tab though

 

[Jens Roos]

Is your power management setting set to "prefer maximum performance"?
Check your gpu clock speeds. My 1070 dropped into 2d mode with rF2 and AMS, resulting in stutter although in theory, it would've been fine.

I have maximum performance and pre-render to 1 (and tested with 2 as well, seems to be recommended for SLI). Run 95% of all games full screen, have similar experience as you when it come to borderless/windowed.

What I mean with "low gpu load stutters" is that, when in traffic or a part of the track with "heavy" graphical load, it run smooth and as it should. When there is no other visible cars you get micro stutters the gives an visual of a bad slowmotion (no tearing), but it's obvious not. So far, only by setting sync to video makes it smooth. This only happens in rF2.

I have previously monitored the clock speeds, bus usage, power usage and nothing out of the ordinary (then at least). What exactly to you mean with 2D mode?

 

[RasmusP]

I have maximum performance and pre-render to 1 (and tested with 2 as well, seems to be recommended for SLI). Run 95% of all games full screen, have similar experience as you when it come to borderless/windowed.

What I mean with "low gpu load stutters" is that, when in traffic or a part of the track with "heavy" graphical load, it run smooth and as it should. When there is no other visible cars you get micro stutters the gives an visual of a bad slowmotion (no tearing), but it's obvious not. So far, only by setting sync to video makes it smooth. This only happens in rF2.

I have previously monitored the clock speeds, bus usage, power usage and nothing out of the ordinary (then at least). What exactly to you mean with 2D mode?

Before the "optimum performance" setting there was only "adaptive" which did split the clock speeds of my GTX 970 into 3 areas: (can't remember them all so I'm gonna use the ones for my 1070):
139 MHz: Desktop mode
139-1582 MHz: somewhere between these, mostly around 800 MHz, watching youtube videos, browser games etc
1582-2012 MHz: 3D mode for gaming. With "prefer maximum performance" it won't go below 1582 MHz anymore

Now the first 2 ranges I call "2D mode" and only the last range I call "3D mode".
Now if a game has such a low GPU usage that the 2D mode would be fine, with "adaptive" or "optimum" it will try to run games like Automobilista at 800 MHz.
Stutters like crazy, clock speed jumping around etc.

"So far, only by setting sync to video makes it smooth. This only happens in rF2."

Kinda strange but what happens when you set sync to video but limit the fps 3 fps lower than the Hz setting? Gsync overrules vsync so you shouldn't experience any input or stuttering then, since vsync wouldn't be active. Maybe it helps with SLI in rF2..

 

[Jens Roos]

"So far, only by setting sync to video makes it smooth. This only happens in rF2."

Kinda strange but what happens when you set sync to video but limit the fps 3 fps lower than the Hz setting? Gsync overrules vsync so you shouldn't experience any input or stuttering then, since vsync wouldn't be active. Maybe it helps with SLI in rF2..

So so far I have this:
vsync off, gsync on, fps limit 101, monitor 100Hz - tearing (stutter unknown)
vsync off, gsync on, fps limit 101, monitor 120Hz - no tearing, low gpu load stutter
vsync on, gsync on, fps limit 101, monitor 100Hz - no tearing, no stutter

So left to try is:
vsync on,gsync on, fps limit 101, monitor 120Hz - i try this later today

 

[RasmusP]

So so far I have this:
vsync off, gsync on, fps limit 101, monitor 100Hz - tearing (stutter unknown)
vsync off, gsync on, fps limit 101, monitor 120Hz - no tearing, low gpu load stutter
vsync on, gsync on, fps limit 101, monitor 100Hz - no tearing, no stutter

So left to try is:
vsync on,gsync on, fps limit 101, monitor 120Hz - i try this later today

I would suggest:
vsync on, gsync on, fps limit 97, monitor 100 Hz

For my 3440x1440 Monitor, 120 Hz is "overclocking area" but it's also beyond the limit for the DisplayPort 1.2 spec sheet as far as I know. DP 1.2 ends at 101 Hz for that resolution!
So maybe the critical thing is 120 Hz and not vsync, gsync etc?

My comments would be like this:
vsync off, gsync on, fps limit 101, monitor 100Hz - gsync off due to fps beyond Hz

vsync off, gsync on, fps limit 101, monitor 120Hz - gsync on but problems with such high Hz setting at your resolution
vsync on, gsync on, fps limit 101, monitor 100Hz - gsync inactive, vsync active since fps hitting max Hz, which deactivates gsync

 

[Jens Roos]

My monitors are 144Hz, and can be "overclocked" to 165Hz.
My gpu:s has DP 1.4, but the monitor has DP 1.2

But still, it should be "no problem" running 2560x1440p @ 165Hz with DP 1.2

https://en.wikipedia.org/wiki/DisplayPort

The 120Hz limit is in the nvidia surround @ 7680x1440 resolution.
But yes, the new wider screens can run into Hz limits with these bigger resolutions.