Nvidia Inspector introduction and Guide

nvidia inspector

Enabled fast sync but somehow in nvidia inspector i havnt got a 144hz Preferred refreshrate option only max 120hz, yet my monitor is 144hz any ideas?:3eyes:

 

Idea: Post in the regarding nvidia inspector thread!? :bang:

 

what games u playing?

 

From the top post:

I downloaded the latest profile inspector, and I saw it has new frame limiter values (labelled "v2"). With that, there is zero latency with 1/2 refresh (well, apart from the natural frame latency of the specific frame rate you're getting.)

I went further and used 1/3 sync to make it even worse. In 120Hz, this results in 40FPS in Doom 4 (Vulkan renderer). However, setting the "v2" frame limiter to 40FPS results in zero input lag. This feels quite unreal. There you get 40FPS with vsync, yet the mouse is extremely snappy and low-latency.

This is very weird. I can't find much information about this "v2" limiter. What does it do and why does it eliminate input lag so well? If I disable the limiter, the mouse in Doom 4 at 40FPS feels like a boat.

Edit: No g-sync or freesync involved here. Just plain ol' vsync. I'm on a 980 Ti and my monitor does NOT have gsync.

 

Yep, fps-limiter v2 seems to be running in coordination with vsync if you use values that are close to display refreshrate or to deviders of it.

Example:
My display is running with 75Hz. If I enable vsync and set fps-limiter v2 to either 37 or 38 fps, RTSS overlay reports a framerate of 37.5fps and there is zero stutter.
(not sure about the input lag though, feels like normal DB vsync for me)

Edit: Happens also without vsync, by just setting fps-limit to devider fps.

Funny thing: The devider vsync options (e.g. 1/2 refreshrate) don't work for me at all in Talos Principle, when it is set the game still runs with synced 75fps.
So the fps-limiter v2 can be used as a workaround to achieve this behavior despite of the option meant for it being broken.

 

Whoah, I will have to try this out. As is currently, without changing the pre-rendered settings 1/2vsync has very high latency. I remember when I played Mafia II on my 570, I kept running into microstutter when playing at 60hz sync but with a 30FPS cap. So I had to use 1/2 sync but it made a lot of actions like aiming very difficult and dare to say. Significantly laggy. (With a controller no less). I didn't know about the after mentioned trick at all. And when I went back and tested it later, the difference was very big.

I haven't done much if any experimenting with the built in Frame rate limiter since all the changes have happened to it because my experiences with the original were fairly poor.

But I will have to find some time to play around with this and add information to the OP.

 

If you set 1/2 vsync, then input lag is waaaay higher than normal DB vsync. And that's with pre-rendered frames set to 1, even. It's over 100ms. The "100ms" number is a guess; I don't have a timer in my brain . But I can definitely see that some time is passing between me moving the mouse and the view changing on the screen, and that time seems to be somewhere between 0.1 and 0.2 seconds. The frame limiter brings the input lag back to what DB vsync should be, to levels where I can't actually even try to measure the lag. I can tell some lag is there, but it's so small that counting tenths of a second is impossible. So from that I can assume it's not even remotely close to 100ms.

Now for 75Hz, you can set the frame limiter to 38 without 1/2 sync and it should work OK, since at that frame rate the microstutter is not very noticeable. But if you do 120Hz with a 60FPS or 144Hz with a 72FPS cap, then the microstutter *is* very noticeable. 1/2 sync is the only way to get it smooth without microstutter. But that seems to add huge amounts of input lag without the v2 limiter.

Doing some further tests, the normal, non-v2 limiter also seems to do the same thing now, but it's more hit and miss when it comes to micro-stutter. But again, maybe there's some placebo going on here. Unfortunately, I don't have any kind of FCAT gear that would allow for actual measurements.

It would be a very interesting article if Guru3D would meassure these things. Do FCAT measurements with RTSS, driver-level limiter, driver-level v2 limiter and perhaps also with 1/2 and 1/3 sync thrown in.

 

I remember it like this as well. But before I can refresh my memory, I need to find a game where it actually works. It does not in Talos Principle. It does just normal vsync.

There is no microstutter at all, because it perfectly renders with 1/2 refreshrate.
RTSS reports 37.5fps (= 1/2 of 75). Frame limiter v2 usually never caps after decimal place.
I'm very certain that there are no repeated frames aka stutter.

Hm, weird. Why would it work fine with 75Hz, but not other refreshrates?

 

Did some further tests, and it works differently with different games. For example, with No Man's Sky, the frame cap is off a bit. If I set 60, I get 59.5, which has one frameskip about every two seconds or so. With Dirt Rally, the cap is dead-on. If I set 60, it's 60. Same with Elite Dangerous and Obduction. With Doom 4 OpenGL, 60 caps to 59.7. With Doom 4 Vulkan, 60 caps to 59.5.

Well, this might also depend on the game then. With the games I tried this on, capping to half refresh in 60Hz (so 30FPS) without 1/2 sync does add some microstutters, but I have to look very, very closely to notice them, since the frame rate is already low enough to be naturally "stuttery" on its on. Any small deviations in FPS simply get lost in the noise. 37.5 is also low enough for this to be true.

If I cap to 60FPS in 120Hz without 1/2 sync, micro stutters stick out way more, since 60FPS is smooth enough to notice even small deviations.

 

Update 1.9.7.7

 

Unreal, so you depended on this file for some stuff as well?
That's irritating.

 

I did try 1/2 refresh sync with 30 fps ( Frame Rate Limiter v2 ) and pre-rendered frames 1 vs without the limiter. And surprisingly I could feel a difference in the game I tested (Dead Rising)

I tried the 30 V2 limit with 60hz sync and input lag actually felt about the same as 1/2 sync with pre-render 1. I have a feeling that's just down to this game though.

 

NVProfileInspector 2.1.2.15 and 372.70 and 372.90 can't accept frame rate limiter in the global profile.

Every time I restart the machine the limit in the global profile changes to "off", only the limit in especific games profiles keep. Anyone else having this issue?

 

2.1.3.0 is out. no such issue so try to clean install again.

 

Did anyone get the new (ver) 1.9.7.7 http://www.softpedia.com/get/System/System-Info/NVIDIA-Inspector.shtml i installed it last night works good

 

Uninstall Geforce Experience, it resets settings on restart.

 

But be careful when setting fps-limit globally, it can slow down programs with accelerated UI like Firefox (even when fps-limit is above refreshrate).

 

Thank you, it did the trick. I usually do not install GFE, but, with the new version I resolved to look at it again, but, that think have no use to me.

Slowdown in what way exactly?

 

Firefox scrolling gets stuttery and even moving its window can be stuttery (more than normally).

 

so what are those new frame rate limiter options? are they also available w/o inspector in vanilla nvcp?
op can you update the main post?

edit: btw in my experience, rtss's frame rate limiter worked better in all situations. everytime.