flood's input lag measurements

[flood]

I tried playing for about a month with 340.43, but all I had - latest place and no aim. I've even lost 3 rank positions. After installing 340.52 - magic happened. Mouse feels ok again and I brought back my previous results like 1-2 place in competetive mode and guardian elite - distinguished guardian rank.

you are aware that like a month or two months ago there was a huge wave of hackers around those ranks... all of my friends (~dmg) deranked.

i don't think what you and those other people experience is actually input lag... it could be like microstuttering or something

here's a reference for what additional input lag feels like:

go to the windows desktop
move the mouse around at the top of the screen
move the mouse around at the bottom of the screen

feel the difference? probably you don't, but at the bottom of the screen there is an additional refresh cycle's worth of input lag. i.e. 16ms for 60hz, 8ms for 120hz.

[Dragonheart]

I tried playing for about a month with 340.43, but all I had - latest place and no aim. I've even lost 3 rank positions. After installing 340.52 - magic happened. Mouse feels ok again and I brought back my previous results like 1-2 place in competetive mode and guardian elite - distinguished guardian rank.

you are aware that like a month or two months ago there was a huge wave of hackers around those ranks... all of my friends (~dmg) deranked.

i don't think what you and those other people experience is actually input lag... it could be like microstuttering or something

here's a reference for what additional input lag feels like:

go to the windows desktop
move the mouse around at the top of the screen
move the mouse around at the bottom of the screen

feel the difference? probably you don't, but at the bottom of the screen there is an additional refresh cycle's worth of input lag. i.e. 16ms for 60hz, 8ms for 120hz.

I was deranked simply by being the worst and constantly losing with my team as a result

Nothing like microstuttering happens, gameplay is completely smooth. Small microstuttering and hitches only happens during the first minutes of the game.

I'm pretty experienced in UT2004, actively played it for about 7 years, participated in some tourneys, so I really know how the input lag feels, no offense BTW, I've asked if you have Creative X-Fi because I found out that in Gaming Mode with all the features on it gives me additional lag in ut2004 with openAL + HW/EAX. Some time I had to switch to integrated realtek before playing the game every time untill I found out that X-Fi in Audio Creation mode reduces this lag almost completely, being close or the same as realtek chip. I couldn't explain it, but I found 2 more experienced ut2k4 players who had the same problem.

Any change in input lag is esasy to feel in dynamic play speed (smth like 21:20) and it's impossible to keep the same results with it.

P.S. Can't feel any difference between the top/bottom of the screen in 144hz mode

[flood]

the only sound stuff i have is onboard realtek and an hrt microstreamer dac. neither have any issues for me, though i am one of those people who have never noticed any changes in feel from driver versions, bios settings, etc...

the hackers and whatever don't just affect you directly. what happens is that people in a higher actual rank than you get deranked from playing hackers. then when those players play with you, you have a harder time keeping up with them and bottom frag. anyway that's something else to think about.

that test isn't really perfect since people are completely used to that little effect. maybe try this: http://forums.blurbusters.com/viewtopic.php?f=10&t=1134

if it really is just simply increased input lag, it can't just be like 3 4 or 5ms if it affects performance so drastically. more than one pro csgo lan tournament has been held using eizo fg2421s which are measured to have around 10ms additional input lag compared to the TN 24in 144hz monitors. and no one complained about lag...

are you aware of noacc's tests?
http://esreality.com/post/2640619/input ... -csgo-q3a/

Things, that do not affect input lag (or their impact is less than 2-3ms). 10 tests for all
1. Maximum prerendered frames setting in nvidia control panel both for QL/CSGO
2. Scaling setting in nvidia control panel. QL
3. Ingame resolution. QL
4. Turning on/off aero. QL
5. GPU scaling. QL
6. Monitor scaling. QL
7. Turbo Boost/C-States/EIST/Thermal Monitor/Etc. QL
8. PEG Gen2/Gen3 Slot Configuration. QL
9. Execute Disable Bit. QL
10. Intel Virtualization Technology. QL
11. xHCI pre-boot driver / xHCI mode / xHCI hand-off. QL
12. eHCI hand-off. QL
13. On-board audio (:D). QL
14. Wake on Lan (:D). QL
15. Java, Flash, IE, Print Spooler, Human Interface Device Access (:D:D:D). QL
16. Turning HPET ON/OFF with command line in Windows 7, while HPET is ON in BIOS. QL/CSGO
17. BenQ overdrive options. QL

Things, that do affect input lag
1. 32ms avg for 125fps (30 measurements). QL
2. HPET OFF = stutters + drops in frame rate + it adds HUGE input lag variace - I made 10 measurements and it was anything from 26ms to 90ms (!!!) turning HPET on/off with cmd in WIndows 7 did not make any difference QL/CSGO
3. Lightboost tends to add couple ms, but since this is within measurement tolerance and I actually need to perform 50+ events checking I got lazy about doing more measurements. QL
4. BenQ built in featrues Aspect, 17" and 19" are all adding 5-8ms. Take a note that those are only available with 120Hz, so it might be partially due to the lower refresh rate. QL
5. Turning off scaling completely in nvidia control panel resulted in ~3ms more input lag, but it might be due to the low amount of measurements. I will perform more for this later. QL
6. Instant mode OFF = +7ms lag to the color state, but grey reponse is still ~28ms QL
Picture for Zowie EC2 CL

though due to his testing methodology the variance between measurements depends a lot on the refresh rate, so it's a bit harder to draw solid conclusions.

[Dragonheart]

if it really is just simply increased input lag, it can't just be like 3 4 or 5ms if it affects performance so drastically. more than one pro csgo lan tournament has been held using eizo fg2421s which are measured to have around 10ms additional input lag compared to the TN 24in 144hz monitors. and no one complained about lag...

Well, that's not the best example I remember the times when LCDs just came out and gamers of every pro lan tournament were forced to play on this laggy shi~ with 60hz. No one was glad about it after playing on fast CRT for years, but who cares - it's all about money. Same goes to the eizo - no sense of complaining about the hardware that sponsors provide.
Also, these cybernerds have better adoptation abilities and since they are playing 24/7 it's easier to somehow adjust and compensate some small increase in delay. Brain's abilities are infinite
I myself had a desire of purchasing this eizo model since VA is sweet, but it's lag measurements were the reason not to do that
I also agree that such a small increase in input lag may be unnoticable, but in case of these drivers I feel like it's much higher than 3-5ms.

are you aware of noacc's tests?
http://esreality.com/post/2640619/input ... -csgo-q3a/

though due to his testing methodology the variance between measurements depends a lot on the refresh rate, so it's a bit harder to draw solid conclusions.

Yeah, I've read it right before I found that thread I've even asked in his topic about testing these drivers, but it seems like there's no activicy from him lately.
The only thing I can't agree is HPET which not always works as it should thus giving an increased lag on some systems - the reason why there's so many different views on it over the web.

[flood]

yea it's really hard to disprove or prove anything because some of these issues may be very hardware and or game specific. my current mobo (asus z87 pro v edition) has no option to disable or enable hpet unfortunately. but i can try it on my old i7 920/x58 desktop some time

[flood]

nvidia gtx 970 on 344.11, win 7 64 bit
in game fps ~1600 (really simple custom map)

3 3 5 2 3 4ms

these are significantly lower than my previous 5-6ms measurements... o.O


gtx 970 doesn't have older drivers so i'm gonna swap it with my gtx 460 and try
well actually im gunna return this gtx970 anyway because it sucks lol

[flood]

forgot to mention that previous post was for csgo
gtx 460 @ 344.11 w7 64bit
framerate ~1000
6 3 3 4 ms


this is strange, before i would always get 5-6ms...
could be win8 vs win7... if so then i may have just discovered something very important

but i'm really happy to see these numbers as it is near the limit for what is possible with a 500hz mouse and a 1000fps game.

[flood]

340.43
6 7 4 4 4 5 4 3 5 3 5 5 3ms

340.52
6 5 3 5 ms

all i can say at this point is that for my setup on a local server with ingame fps of ~1000, i so far have not found a significant difference in average input lag between the driver versions.

it could be that one of the driver versions has an additional frame or two of buffering, which would not really be noticeable in these tests since at 1000fps as it would just be 1/2ms of delay. but at lower framerates it could be a factor

[Dragonheart]

Thanks a lot for the results, flood!

It would be really interesting btw to see how much of advantage in lag 1000/1600fps give comparing to smth like 300fps.

[flood]

more csgo measurements
same settings but varying fps_max
no cap (~900fps stable): 5 6 5 6 5
500: 5 4 5 7 4
250: 7 5 5 7 7
125: 8 8 6 5 10 8 8 5 11 11

what happens when the framerate is actually low instead of just being capped low?

i changed the map to cs_insertion which is a more graphically intensive map. the fps for these were all 300 +- 20

9 11 12 12 11 9 11 9 12 13

this matches with what i expect:

input lag = base input lag + frame render time + X

base input lag would be the input lag with infinite framerate
X is a random variable between 0 and the "time between frame"

fps_max reduces the time between frames, but doesn't affect the render time. that's why the fps_max measurements were sometimes very low, even at fps_max 125.

just so i'm following on one of your lines of thought:

assuming equal base input lag, over many measurements one would expect the same distribution of input lag in both these situations:

1) artificially capping a game at 125 fps (when uncapped it could do 1000000 fps)
2) a game that can only render at 125 fps even when uncapped

correct?

suppose the base input lag is 0 (i.e. uncapped has 0 input lag
for 1 you'd get anything between 0-8ms lag
for 2 you'd get anything between 8-16ms lag

this depends on the cap working by sleeping after drawing instead of sleeping between getting the input and drawing.

explanation for my formula: http://i.imgur.com/9cSP1bM.png