ESR - Frame metering, i need your help.

<< Comment #1 @ 12:28 CDT, 12 August 2014 >>

(Link, Reply)

By

doped

so no one gives a shit? or ppl who did moved to ati?

<< Comment #2 @ 13:22 CDT, 13 August 2014 >>

(Link, Reply)

By

horf - Reply to #1

A bit of both in my case

<< Comment #3 @ 13:53 CDT, 13 August 2014 >>

(Link, Reply)

By

solobeNNN

luckily i use AMD and use radeon pro to set the rendered frames to 0 and gg no re no input lag

<< Comment #4 @ 15:33 CDT, 13 August 2014 >>

(Link, Reply)

By

doped - Reply to #3

prerendered frames have nothing to do with input lag though imo, it only has effect when you get slight frame hitching in a demanding game. I could be wrong though. All that dpc latency is completely bull. Problem is added delays in subsystems to control framepacing, even if it's only said yo be used when using sli, it's still in full effect all the time as i see it. Even if we got an option to disable it, would we still have the added delays from various tasks still?

Nvidia don't give a shit, as long as they have "smooth" frames in benchmarks, which they provided the tools for.

I'm amazed no-one else gives a shit, or did i miss the storm? am i too late in trying to draw attention to nvidia's shitty drivers being unusable for anything where linear/undelayed mousefeel in very important?

<< Comment #9 @ 03:46 CDT, 14 August 2014 >>

(Link, Reply)

By

noacc - Reply to #4

prerendered frames have everything to do with input lag actually. Added input lag by this feature = number of those frames. If you have 125 FPS and enforce prerendered frames to 3 in nvidia control panel you will get 24ms added input lag. And you will notice it in game immideately. Setting it to one and max FPS to 250 virtaully eliminate that lag, as ya are only one frame (4ms) behind the game world.

<< Comment #10 @ 04:40 CDT, 14 August 2014 >>

(Link, Reply)

By

Emenike - Reply to #9

So you wanna say you did test it at Quake Live?

<< Comment #30 @ 15:41 CDT, 14 August 2014 >>

(Link, Reply)

By

noacc - Reply to #10

yes, you can test yourself and notice added lag and some glitches especially when you are suddenly gaining huge amounts of speed like with good CJ.

<< Comment #11 @ 07:55 CDT, 14 August 2014 >>

(Link, Reply)

By

doped - Reply to #9

it's MAX, not always, i'm not saying I understand how openGL rendering works, but i don't think that's how it works, I've never had input lag on old drivers, even with pre rendered frames on 3. Afaik, it has to do with framehitching, if there's a hitch, the following frames can be made to smooth out that hitch, so every consecutive frame is not affected, if only 1-2frames was the problem.

pre-rendered frames are not designed, or supposed to have the gpu make 3 frames, THEN push them to display, that's what vsync buffer is for.

<< Comment #31 @ 15:57 CDT, 14 August 2014 >>

(Link, Reply)

By

noacc - Reply to #11

first, i did answe to this
prerendered frames have nothing to do with input lag though imo

not to your original post (because it is usless to argue without experimental data).

second. Read its description please:
Maximum Pre-Rendered Frames: If available, this option - previously known as 'Max Frames to Render Ahead' - controls the number of frames the CPU prepares in advanced of being rendered by the GPU. The default value is 3 - higher values tend to result in smoother but more laggy gameplay, while lower values can help reduce mouse and keyboard lag.

So CPU is preparing 3 frames in advance for GPU. What does this mean? It means that when you see the oldest frame on your monitor it will be already behind actual game time by 3 frames (because while GPU is rendering and display is actually drawing the oldest frame CPU already calculated 4th frame which represents current game time.

Anyways I am interested in actual full input lag measurement. Getting 1000fps camera this weekend and going to mod my mouse with led tomorrow. So if you want me to test full input lag difference with different versions of drivers I can do this. However my GPU is modern 760 hawk, so I won't be able to test very old drivers I guess. I also prolly borrow g520 crt again just to make a test.

Edited by noacc at 16:23 CDT, 14 August 2014

<< Comment #33 @ 17:06 CDT, 14 August 2014 >>

(Link, Reply)

By

earth - Reply to #31

Now that is interesting :D Please post a slow-mo video of a CJ or something and what you find.

<< Comment #39 @ 00:56 CDT, 15 August 2014 >>

(Link, Reply)

By

noacc - Reply to #33

thats impossible, the camera I am getting can only shoot high fps on very low resolution. Also you can google "strafe jump tutorial" there are videos with 0.1 timescale or even less with key visualisation )

<< Comment #34 @ 17:21 CDT, 14 August 2014 >>

(Link, Reply)

By

doped - Reply to #31

I stand corrected, I always thought it was a when-if, thing, and not all the time, but thanks for clearing that up. I set it to 1 now, dno if it helps

<< Comment #40 @ 00:58 CDT, 15 August 2014 >>

(Link, Reply)

By

noacc - Reply to #34

well, on 250FPS it will not be that much noticeable because frametime itself is very low. However on 125 it's effect is very obvious.

<< Comment #42 @ 07:26 CDT, 15 August 2014 >>

(Link, Reply)

By

Anonymous (78.241.13.74) - Reply to #9

it has been proved many times already that it does not work this way. i myself tried 5 prerendered frames with a 125 fps QL on a 200hz CRT nad it had ABSOLUTLY no more input lag than prerendered frames 0/1.

14%

<< Comment #43 @ 15:48 CDT, 15 August 2014 >>

(Link, Reply)

By

noacc - Reply to #42

ok.

I finally got my hands on 1000fps camera and modded my zowie ec2 evo cl with led which lights when mouse1 is pressed.

My PC specs are i5 4690 everything at default in bios and OS, gtx 760 hawk everything set to highest performance and lowest quality, win 7 64bit with latest updates, latest nvidia driver. Game running at 250FPS, monitor is xl2411t at 144hz.

Average overall input lag measured from led lit to rail trail appearance on screen is 28ms with variance from 24ms to 32ms.

So this lag did NOT depend in any way of:
1. maximum prerendered frames hence you are absolutely correct anon.
2. GPU/monitor scaling.
3. Resolution used.

Also my friend kindly let me measure input lag using G520 CRT. At 144Hz I could not find measurable difference in input lag. At 170Hz input lag tended to be couple ms less. I did not get anything more than 29ms. Lowest was 20ms.

I did perform ~50 measurements for LCD and ~20 measurements for CRT (got lazy with that one :-) )

125fps vs 250fps was 32ms average.

Measurement tolerance is estimated at �2ms.

Also this xl2411t appeared to have some tricky algoprythm of brightness adjustment. From 95% at OSD it seems like it completely eliminate blanking interval between the frames or at least I can not catch it with 1000fps camera. At 100% brightness this monitor looks like still image at 1000FPS. I will upload the footage when I get usb cable for the camera.

<< Comment #45 @ 07:16 CDT, 16 August 2014 >>

(Link, Reply)

By

doped - Reply to #43

you should definitely make a separate post with you finding, would it be possible to test ati cards too, that would still be interesting.

<< Comment #5 @ 17:48 CDT, 13 August 2014 >>

(Link, Reply)

By

byce

I'll admit I didn't read your post word-for-word but I don't think you mentioned what card you have. With an older card you probably will need to stick with an older driver. I think I had to use the same driver you're stuck on with my old 9600gt. Also if you're on XP you need com_idlesleep=0. I know that's probably obvious stuff but someone has to say it so we can get it out of the way.

Edited by byce at 17:49 CDT, 13 August 2014

<< Comment #12 @ 07:57 CDT, 14 August 2014 >>

(Link, Reply)

By

doped - Reply to #5

it's out of the way now, i'm just trying windows 7 for this very reason, to see if nvida is neglecting xp with a huge bug like this, and i can report, nope, not a bug, it was designed this way. Frame pacing is still shitting all over mouseinput even at stable 125fps.

<< Comment #6 @ 20:32 CDT, 13 August 2014 >>

(Link, Reply)

By

Anonymous (84.238.88.67)

I would like to know if the response time is the same on all AMD cards/drivers? Or did they not fuckup like NVIDIA did? all 300.xx drivers are so horrid.

<< Comment #13 @ 07:59 CDT, 14 August 2014 >>

(Link, Reply)

By

doped - Reply to #6

im a bit hard pressed here, but latest driver for hd4850(which is not latest from amd, since that card is moved to legacy), 13.9 is lag free like old nvidia drivers, and ati has always been.

word of advice though, although they are lag free, if you care at all about old games, you are out of luck, ati really don't give a shit if they break only years old games. Latest driver have srs problems with spawning particles in older opengl games, and bogs down like it's and ati rage card. That's for 13.9, they might have fixed it since nvidia had the same problem at one point, i think it's related to a specific version of newer opelgl, and it got fixed.

Edited by doped at 08:08 CDT, 14 August 2014

<< Comment #7 @ 23:05 CDT, 13 August 2014 >>

(Link, Reply)

By

odellus

go tell mark rejhon about it and see if he can do some tests or something. i wasn't into perfecting my system back then so i don't really know if what you're saying is true and i don't have a way of testing it because drivers that old don't support my graphics card. you need proof.

<< Comment #14 @ 08:04 CDT, 14 August 2014 >>

(Link, Reply)

By

doped - Reply to #7

i can only prove it with frame metering being introduced at that time, and the fact that i come from 1999 with crt's and the works. I know how a mouse is supposed to respond, and when changing from old driver to newer, they are so much off on the mouse feeling. It's not a placebo, lots of other people agree. I didn't know other people had the same problems until recently, i have stuck with the old driver for so long now, exactly because feel was off, but now im trying to do something.

mark rejhon, is he from techreport?

<< Comment #16 @ 08:11 CDT, 14 August 2014 >>

(Link, Reply)

By

odellus - Reply to #14

he's the blurbusters guy. i'm not saying you're wrong, in fact i'd like for you to be right because that'd mean there are improvements to be made, but you have to have solid proof, not anecdotes.

Edited by odellus at 08:12 CDT, 14 August 2014

<< Comment #18 @ 08:13 CDT, 14 August 2014 >>

(Link, Reply)

By

doped - Reply to #16

that will come, im on their site aswell, trying to get answers.

<< Comment #17 @ 08:12 CDT, 14 August 2014 >>

(Link, Reply)

By

doped - Reply to #7

can't you inf mod your card into using an older driver? you should give it a try at least.

<< Comment #8 @ 00:31 CDT, 14 August 2014 >>

(Link, Reply)

By

Emenike

What videocard and OS do you have? May be only you have this problems?

<< Comment #15 @ 08:05 CDT, 14 August 2014 >>

(Link, Reply)

By

doped - Reply to #8

not at all, just do a google search, you'll find lots of people of the nvidia forums, overclock forums, and the cs community have guides how to supposedly get rid of it. Obv. they don't work, as it's deep in how the driver works.

<< Comment #19 @ 10:04 CDT, 14 August 2014 >>

(Link, Reply)

By

Mental Patience - Reply to #15

I Googled and just found many r0ach posts insulting people for being inferior.

set your monitor to not scale if lcd, 1:1 in osd.
set : do not scale in nvidia control panel.
set : perform scale on monitor in scale options.. not gpu.
pre render 0 if frame rate higher than refresh.
try disable thread optimization. (old games).
make sure you have full hz by checking osd.(59 119 bug )
remove all nvidia 3d & physix bloat for twitch games also fixes 59 Hz bug.
still no good? get someone to prove it with high speed vidya, then show nvidia?

Edited by Mental Patience at 10:29 CDT, 14 August 2014

8%

<< Comment #20 @ 10:46 CDT, 14 August 2014 >>

(Link, Reply)

By

doped - Reply to #19

im sorry, but ive tried all that already. Roach is an idiot by the way, he's right about drivers being in the shitter, but his dpc latency, uninstall flash and java bs is such a load of crap, and what else good ideas he's come up with, the man is desperate, and i see how, but he's on a wrong path, he should investigate, instead of trying to fix broken drivers by doing all those other things.

It's the nvidia drivers, and you can do nothing about it. Now, how am i going to show inconsistent mouse feel with a camera, caused by something messing with input? I'm interested in finding out how i can prove my claims. I'm hesitant to call it input lag, and that's why it's so hard to prove. If it was a consistent input lag, you could point a 1000fps camera on screen showing both physical mouse movement, and onscreen movement, easy.

Also, i get a strong vibe you do not believe me, what's in it for you to defend nvidia? what if im right, and it's you, who do not feel it? I'm hesitant to attack you not feeling it, but have you've ever played on a crt + pre-shitty drivers and know how the mouse is supposed to feel? It's a problem since 2008 and that's a big problem for guys like you not remembering how mousing has to feel.

<< Comment #21 @ 12:00 CDT, 14 August 2014 >>

(Link, Reply)

By

Mental Patience - Reply to #20

just mentioned my settings to try and help.
I think most people are happy with gpu scaling off, that's the problem.
if you could raise this issue with nv and get an option to improve mouse response and input lag then hats off to you.
I can't test this and prove anything.
maybe a high speed camera could, if the mouse and cursor or clicks with both drivers was examined? it seems the only way to measure IL.
most people describe the issue as IL.
crt, nv, pre 190... yes, I must have gotten used to it over the last 6 years and forgotten how good it was.

I'm not defending nv, red or green, I don't care, I buy both ;

Edited by Mental Patience at 17:47 CDT, 14 August 2014

<< Comment #27 @ 13:45 CDT, 14 August 2014 >>

(Link, Reply)

By

doped - Reply to #21

I appreciate you trying to help.

<< Comment #23 @ 12:09 CDT, 14 August 2014 >>

(Link, Reply)

By

Emenike - Reply to #15

I see dumb prople IRL everyday, no need to google them. Why are you dodging answer question? May be coz you are using old videocard at old OS like windows xp?
It's ok for monopolist to abandon old their product support and no need to be surprised if nvidia don't give a fuck about input lag at old videocard with old operating system.

<< Comment #26 @ 13:43 CDT, 14 August 2014 >>

(Link, Reply)

By

doped - Reply to #23

what are you trying to say? what did i dodge? it has nothing to do with what video card you use obv. and im on windows 7 now, so no, im not dodging your questions. Anything more you wanna ask us dumb ppl?

<< Comment #29 @ 15:07 CDT, 14 August 2014 >>

(Link, Reply)

By

Emenike - Reply to #26

I wanna say your problem with outdated windows xp, old nvidia card and newer drivers has chances to be real. But that doesn't mean that nvidia have input lag drivers. Only in your case may be yes.
You know supporting old products is kinda useless.

4%

<< Comment #35 @ 17:26 CDT, 14 August 2014 >>

(Link, Reply)

By

doped - Reply to #29

I'll come back to you, because i upgraded to windows 7 yesterday, and there's a difference so far, if it sticks a couple of reboots, i might have found a solution, and it is idd a bug with windows xp. I'm not saying it just yet, but so far after installing win7 + 340.52, disabling physx, vsync, cuda, prerendered frames to 1, it was still not right feeling. I just managed to OC mouse to 500hz, and NOW we're talking, there's a difference now, to the better, feels MUCH more linear as it's supposed to, but im not calling it just yet.

Just to let you know it might be an xp bug, and win7 125hz mouse bug.

<< Comment #48 @ 05:29 CDT, 17 August 2014 >>

(Link, Reply)

By

Mental Patience - Reply to #35

wow, you are special needs.

<< Comment #49 @ 06:36 CDT, 17 August 2014 >>

(Link, Reply)

By

doped - Reply to #48

lets see you stay with windows 7 in 5 years when "windows 9 is bullshit" and something breaks and no one cares, but your beloved "windows 7 just works" and that windows 10, with it's fancy fisherprice touchscreen friendly whatever they come up with.

come back in 10years when you learned to use something because it just works, i don't have time to upgrade, i don't care to upgrade. And no-one told you that, that's why(the os) that another thing broke.. BADLY (how would you know? it's not readily available anywhere is it now?)

That you choose to call that(me) "special needs", just goes to show that you are still young and clueless about choosing other things than using lots of time upgraded something "that works", but you can have it, if it makes you feel better about yourself.

<< Comment #22 @ 12:02 CDT, 14 August 2014 >>

(Link, Reply)

By

sofiene

i have an ATI video card and never managed to have a stable 125 FPS, always have frame drop even if i use a low res/cfg ( currently with 800x600 ) and no bloom effect, always asked for a solution and never got any :/

<< Comment #24 @ 12:11 CDT, 14 August 2014 >>

(Link, Reply)

By

Emenike - Reply to #22

r_enablePostProcess 0
BTW QL performance much more depends on CPU than on GPU.

<< Comment #25 @ 12:23 CDT, 14 August 2014 >>

(Link, Reply)

By

sofiene - Reply to #24

this is the only answer that i always get, i have r_enablePostProcess 0 for ages and it didnt help :/

<< Comment #28 @ 14:57 CDT, 14 August 2014 >>

(Link, Reply)

By

horf - Reply to #22

With ati dynamic lights and some particle effects seem to affect the framerate more than anything with openGL games.

And also when I had the stuff off I seemed to get more stable performance than I get with nvidia.

<< Comment #32 @ 16:10 CDT, 14 August 2014 >>

(Link, Reply)

By

u3a6 - Reply to #22

Perhaps the load that ql puts on the gpu is not enough to make it go full 3D clocks, hence the bad peformance? I play ql with AA forced because that is the only way I can get a smooth experience (full 3D clocks).

<< Comment #38 @ 21:55 CDT, 14 August 2014 >>

(Link, Reply)

By

sofiene - Reply to #32

how can i do that ?

<< Comment #41 @ 03:40 CDT, 15 August 2014 >>

(Link, Reply)

By

u3a6 - Reply to #38

In AMD Catalyst Control Center go to 3D - Application Settings - System Settings - Anti-Aliasing mode - Override application settings. Play around with the Anti-Aliasing settings to find the balance that yields full 3D clocks and stable 125fps (make sure the load is not too high), usually 30% load in the gpu will be enough to ensure smooth operation. Download gpu z go into the sensors tab an check the clocks (make sure the sensor refresh rate is the lowest possible). Play for a moment and come back to see if there are step or spike like fluctuations in the clock speeds, and also if the gpu is not in power saving mode (the clocks must be the full 3D clocks).

<< Comment #46 @ 08:39 CDT, 16 August 2014 >>

(Link, Reply)

By

Mental Patience - Reply to #38

for old opengl games like quake and ati cards I always use r_vertexlight 1.
helps with fps drops.

<< Comment #47 @ 11:15 CDT, 16 August 2014 >>

(Link, Reply)

By

sofiene - Reply to #46

i have r_vertexlight 1 since forever, it didnt made much of a difference when it comes to FPS

<< Comment #36 @ 18:48 CDT, 14 August 2014 >>

(Link, Reply)

By

qwind

I noticed that with my latest nVidia driver my performance in QL has been shit, resulting in irregular mouse movement.

As long as the FPS is high things feel good, but I am getting very significant drops where I really shouldn't be. My config is very performance oriented (I get better FPS with it than with the QL Performance setting), and I get even further FPS drops if I play on higher settings.

I have an AMD 1090T + 660Ti which should be (and was) more than sufficient.

Not sure what it is exactly with QL these days but it just doesn't seem to play nice with anything considering how simplistic it seems it SHOULD be to run.

<< Comment #37 @ 20:42 CDT, 14 August 2014 >>

(Link, Reply)

By

Anonymous (84.238.88.67) - Reply to #36

The latest driverseries 300.xx from NVIDIA are indeed shit, I lose around 8% accuracy with those installed compared to the 200.xx driver series. I cant test older versions since i got a 680gtx, and the 297.16 is the oldest driver INF i can modify to work with this.

So my next purchase will be AMD no doubt, I'm thinking of buying a 290X card as I'm not only playing QL, Anyone have bad experience with those when it comes to reponse? I've got my eyes on the MSI Lightning edition.

Furthermore "pre-rendered frames 0" is not even possiple with the newest drivers series, lowest is "1". And that did create a shitstorm on geforce forums.

<< Comment #44 @ 06:58 CDT, 16 August 2014 >>

(Link, Reply)

By

doped - Reply to #37

what windows are you on? i just upgraded to windows 7 a couple of days ago. Mouse input is different, but after upgrading mouse to 500hz(bit tricky) and disabling physx etc. I can say, it's SOOOO much better than xp with same driver. To me it seems like a neglected bug in xp nvidia drivers above 180.