Additional input lag with new Geforce driver 337.50?

Everything about latency. Tips, testing methods, mouse lag, display lag, game engine lag, network lag, whole input lag chain, VSYNC OFF vs VSYNC ON, and more! Input Lag Articles on Blur Busters.
aufkrawall
Posts: 6
Joined: 08 Apr 2014, 10:40

Re: Additional input lag with new Geforce driver 337.50?

Post by aufkrawall » 13 Apr 2014, 16:38

I can definitely feel the additional input lag caused by alternate frame rendering (SLI/Crossfire, one additional frame per card).
And not just with low fps, triple buffered (or better said triple queued) vsync can be even more a pain, even with 144fps and Hz.
Good example is Team Fortress 2.
I can't feel this sluggishness with new 337.50 driver and single-GPU.

war1
Posts: 29
Joined: 21 Jul 2014, 09:40

Re: Additional input lag with new Geforce driver 337.50?

Post by war1 » 21 Jul 2014, 10:02

ok, im one of those "insane" persons who can feel this 10ms extra input delay. (it's out of the way, continue reading at your own risk)

I tried roach's dpc latency etc. and found it utter bs, it has nothing to do with anything it's PURELY gpu driver related. The last driver without sluggish mouse feel im able to use is 175.19, but that driver is very old, and does not support hw-accel of streaming video for example, so using it on a day to day computer is not the best idea(plus not supporting newer gpu's so you can say it cant fix the problem to "just use older driver")

I was convinced it was a nvidia driver problem, until i tried an ati card. Although not as pronounced, ati have the same kind of problem, and i believe BOTH have something to do with them trying to fix STUTTER. It's years back now 2009 or something.

I would really like an in depth look at this, but im not smart enough to find the culpit, but it IS to do with the gpu drivers.

A good test for anyone able to pick up on very small changes in input lag: Uninstall gpu driver, reboot. Do the figure of 8 with mouse, take good note of the feel of crispness(if on low latency monitor or crt, motionblur is not relevant at all for this test) NOW, reinstall driver(go newest, has sluggish feel for both) when returning to desktop, do figure of 8 again. FOR ME at least, it VERY small difference, i cant see it(look at hand and cursor) but i feel WAY more disconnected from the movement, feels sluggish.

Gaming on newer drivers is impossible for me, i quickly get RSI feeling in my wrist from trying to compensate eye-hand coordination.


I would REALLY like someone like blurbusters, if able, to have an in depth look at this problem, since it is a problem. And it affects everyone, especially if it has something to do with lengthening the frametime or lower mouse input processing in an attempt to fix some other problem(which i believe at this given time to be the problem)

Now, call me insane, because it is getting to my head.

Q83Ia7ta
Posts: 761
Joined: 18 Dec 2013, 09:29

Re: Additional input lag with new Geforce driver 337.50?

Post by Q83Ia7ta » 21 Jul 2014, 10:34

I'm not insane person but I can feel Lightboost input lag which Chief Blur Buster names 4ms and I can't deal with it.
I have 120Hz LCD since 2010 and 144Hz for last 1,5 years. Tried several decent CRTs for last 1,5 years. Also tried 8800GTS series NVIDIA videocard with Windows XP or 7 and several driver versions. Difference between CRT and LCD (both at 144Hz) is so tiny but also noticeable. After that all my play tests I don't feel any additional input lag by driver version at least for OpenGL at high very FPS values.
May be something wrong with your PC or OS?

war1
Posts: 29
Joined: 21 Jul 2014, 09:40

Re: Additional input lag with new Geforce driver 337.50?

Post by war1 » 21 Jul 2014, 12:00

Q83Ia7ta wrote:I'm not insane person but I can feel Lightboost input lag which Chief Blur Buster names 4ms and I can't deal with it.
I have 120Hz LCD since 2010 and 144Hz for last 1,5 years. Tried several decent CRTs for last 1,5 years. Also tried 8800GTS series NVIDIA videocard with Windows XP or 7 and several driver versions. Difference between CRT and LCD (both at 144Hz) is so tiny but also noticeable. After that all my play tests I don't feel any additional input lag by driver version at least for OpenGL at high very FPS values.
May be something wrong with your PC or OS?
what driver, what os? 337.88 winxp still gives me a tiny amount. You can find drivers far worse, but it's still not as crisp as very old drivers. I can change only driver, and feel a difference, so i know that's the problem.

flood
Posts: 929
Joined: 21 Dec 2013, 01:25

Re: Additional input lag with new Geforce driver 337.50?

Post by flood » 22 Jul 2014, 03:49

A good test for anyone able to pick up on very small changes in input lag: Uninstall gpu driver, reboot. Do the figure of 8 with mouse, take good note of the feel of crispness(if on low latency monitor or crt, motionblur is not relevant at all for this test) NOW, reinstall driver...
I know that feel :D

Not that I don't believe your claims, but I really should make an input lag test to truly see the precision with which we can feel input lag. Of course levels of input lag below the distinguishability threshold may still be significant, but it would be nice to have some knowledge of our own abilities.

Fortunately input lag is something that can be evaluated completely objectively via experimentation; hopefully blurbusters or someone else with the tools can investigate whether drivers do affect input lag.

in the meantime, some questions to think about (I'm not challenging your claims):
can you feel the difference in input lag between moving the cursor near the top of the screen vs moving it near the bottom?
with aero/compositing enabled in windows, can you feel the input lag of clicking things? (probably harder to feel than input lag in motion)
with aero/compositing enabled in windows, can you feel the input lag when dragging windows around? The cursor follows the window perfectly, but actually what's going on is that the drawing of the cursor is done differently and lags the same amount as the window's lag.

war1
Posts: 29
Joined: 21 Jul 2014, 09:40

Re: Additional input lag with new Geforce driver 337.50?

Post by war1 » 22 Jul 2014, 07:43

flood wrote:
A good test for anyone able to pick up on very small changes in input lag: Uninstall gpu driver, reboot. Do the figure of 8 with mouse, take good note of the feel of crispness(if on low latency monitor or crt, motionblur is not relevant at all for this test) NOW, reinstall driver...
I know that feel :D

Not that I don't believe your claims, but I really should make an input lag test to truly see the precision with which we can feel input lag. Of course levels of input lag below the distinguishability threshold may still be significant, but it would be nice to have some knowledge of our own abilities.

Fortunately input lag is something that can be evaluated completely objectively via experimentation; hopefully blurbusters or someone else with the tools can investigate whether drivers do affect input lag.

in the meantime, some questions to think about (I'm not challenging your claims):
can you feel the difference in input lag between moving the cursor near the top of the screen vs moving it near the bottom?
with aero/compositing enabled in windows, can you feel the input lag of clicking things? (probably harder to feel than input lag in motion)
with aero/compositing enabled in windows, can you feel the input lag when dragging windows around? The cursor follows the window perfectly, but actually what's going on is that the drawing of the cursor is done differently and lags the same amount as the window's lag.
Im not super human, so from top to bottom of screen i cannot feel :) But im really not sure anymore, im doubting myself sometimes when ppl say it's just me, and nvidia drivers are fine, but others say the same as me :( I installed 340beta yesterday, and im trying to tell myself the different mousefeel is placebo, and not input lag from driver. :/

if someone had the knowledge to find out what the drivers did different than earlier(pre fermi), we might be able to know what's going on, in the mean time ill just loose my mind over this :)

edit: i just reread something you said!
"The cursor follows the window perfectly, but actually what's going on is that the drawing of the cursor is done differently and lags the same amount as the window's lag."

this might have something to do with it? just a thought. im on windows xp though, but who said they couldn't force always on hw-accel of something on desktop? they use it in browsers and video and such.

Q83Ia7ta
Posts: 761
Joined: 18 Dec 2013, 09:29

Re: Additional input lag with new Geforce driver 337.50?

Post by Q83Ia7ta » 22 Jul 2014, 08:36

war1 wrote:what driver, what os? 337.88 winxp still gives me a tiny amount. You can find drivers far worse, but it's still not as crisp as very old drivers. I can change only driver, and feel a difference, so i know that's the problem.
Windows XP with 296.10, Windows 7 32-bit with 2XX drivers (don't remember which ones), Windows 7 64-bit with 3XX drivers.
What OS and driver combinations did you try? And what monitors, mouses and videocards did you try?

war1
Posts: 29
Joined: 21 Jul 2014, 09:40

Re: Additional input lag with new Geforce driver 337.50?

Post by war1 » 22 Jul 2014, 10:36

Q83Ia7ta wrote:
war1 wrote:what driver, what os? 337.88 winxp still gives me a tiny amount. You can find drivers far worse, but it's still not as crisp as very old drivers. I can change only driver, and feel a difference, so i know that's the problem.
Windows XP with 296.10, Windows 7 32-bit with 2XX drivers (don't remember which ones), Windows 7 64-bit with 3XX drivers.
What OS and driver combinations did you try? And what monitors, mouses and videocards did you try?
i only have winxp to test with, 1xx driver give be best input, 2xx is worst on input lag, and 3xx are a bit better. mx500 @ 500hz, 8800gt and a not so good 19" lcd(besides the point, since it's about the delta-value)

war1
Posts: 29
Joined: 21 Jul 2014, 09:40

Re: Additional input lag with new Geforce driver 337.50?

Post by war1 » 03 Aug 2014, 08:34

although not of windows, nvidia or anything, i think i know what is the problem now. http://superuser.com/questions/455498/d ... ering-in-o (dno if im allowed to post links?)

problem is the buffering of mouseinput to disable tearing on desktop, which in turn makes it feel like vsync. NOW, i think my problem is older games use get_cursor position from desktop, and when it's buffered = mouse delay. OR is it disabled with fullscreen? i don't know.

I'm getting closer. Which is good i guess.

flood
Posts: 929
Joined: 21 Dec 2013, 01:25

Re: Additional input lag with new Geforce driver 337.50?

Post by flood » 03 Aug 2014, 13:23

this is my (unverified but probably true) knowledge:

the window desktop's cursor position is determined at the start of each refresh cycle. therefore the cursor has no tearing AND minimal lag at the top of the screen.
by minimal lag, I mean that when the screen updates, it is displaying the latest possible information.

compared to the top, at the bottom of the screen the cursor has slightly less than one frame of lag. that is why I asked you whether you could feel the difference.

the desktop is rendered with the double-buffered aero compositor, which adds two frames of lag to anything that interacts with the mouse. try dragging around any icon on the desktop and you will see. true fullscreen games, such as those which can show tearing when vsync is off, render directly to the display and bypass the aero compositor.

on os x, the cursor is drawn with the desktop. this means that when dragging icons and stuff, the icon will follow the cursor perfectly , but only because they both lag together. well actually last time i checked in the apple store, icons on macbook airs dont follow perfectly but icons on macbook pros do. :shock:

Post Reply