Page 1 of 44

flood's input lag measurements

Posted: 17 Sep 2014, 19:22
by flood
hardware:
Image

software:
http://pastebin.com/DvgkM9Ab

data:
https://docs.google.com/spreadsheets/d/ ... li=1#gid=0

esreality thread:
http://www.esreality.com/post/2691945/m ... surements/
ocn discussion thread:
http://www.overclock.net/t/1529341/the- ... -input-lag


original OP with some camera measurements:
so i got a casio ex-zr700 which can record 1000fps video at 224x64

original LED method attempt:
disassembled logitech g3 + led flashlight + wires from the fan of an intel stock cooler + fingernails as wire strippers + lots of packing tape =
http://i.imgur.com/95tXIWp.jpg
currently the problem is that the switch is rather mushy and I can make the flashlight turn on/off without the switch actually actuating
then i tried to do this with my g100s and managed to destroy the mouse :D
so i got a new g100s and decided to do motion lag measurements by using the dead g100s to hit the new g100s

anyway,
motion lag measurements
(that means smack the mouse with something and see how long it takes for the screen to respond):
1: barebones love2d program that basically draws a big vertical bar where the mouse is. if the mouse's position changes the color changes.

hardware used here:
sony cpd-g520p CRT, 4770k, asus z87-pro, gtx 460, g100s @ 1000dpi, 500hz without driver, white sheet of paper as mousepad
video recorded with casio ex-zr700 @ 1000fps.

software:
windows 8.1
nvidia driver 337.88
love2d with the following code: http://pastebin.com/aa1pgtkp
resolution: 1600x1200@85hz

results:
10 measurements:
3,5,4,5,3,5,3,3,4,5 ms

give or take 1ms for determining when the mouse starts moving.

sample video: https://www.youtube.com/watch?v=IA0T_o9qQVc

2. a csgo local server

same stuff as above but with csgo.

additional info:
opened a really simple custom map. no fps_cap, scoped into a region where the wall meets the sky, set sensitivity to 200, and hidehud 4. all settings verylow/low, 4x msaa, multicore off. the ingame fps was >900. launch options (shouldnt affect anything): -novid -high -tickrate 128 -w 1024 -h 768 -freq 150 -nogammaramp

result: 5,6,6,5,5,5 ms

video: https://www.youtube.com/watch?v=vlHEpju24-Q

3. same as (2) but varying a bunch of ingame settings

results: see my posts on page 3. everything is as expected. there are no settings in csgo to worry about. as long the fps is high, the input lag is low.

todo:
  • 125 vs 500hz mouse
    repeat csgo measurements at lower fps

    monitor comparisons between crt, vg248qe gsync, and my old crappy viewsonic lcd

    test the nvidia driver paranoia with different driver versions, scaling modes, etc..
    test gtx 460 vs gtx 970 vs intel (im gonna laugh reallllly hard if intel has the least lag)

    asus bios mouse cursor :D
    windows mouse cursor
    x11 cursor in linux
    wayland cursor in linux

    barebones opengl program in linux
    some sort of barebones program in windows (would appreciate help here as I really hate using the winapi)

    cs 1.6
    quake live

Re: my ghetto led lag-measuring setup

Posted: 17 Sep 2014, 20:37
by flood
welp

just fried my g100s somehow with a short :/
note to self: dont stick wires in mouse which usb is plugged in

Re: my ghetto led lag-measuring setup

Posted: 18 Sep 2014, 02:12
by spacediver
ahah sorry to hear that :)

lookin forward to your results!

Re: my ghetto led lag-measuring setup

Posted: 19 Sep 2014, 18:32
by flood
alright im gonna do motion based tests by slamming my broken g100s onto my new g100s.

todo:
monitor comparisons between crt, vg248qe gsync, and my old crappy viewsonic lcd

test the nvidia driver paranoia with different driver versions, scaling modes, etc..
test gtx 460 vs gtx 970 vs intel (im gonna laugh reallllly hard if intel has the least lag)

asus bios mouse cursor :D
windows mouse cursor
x11 cursor in linux
wayland cursor in linux

barebones opengl program in linux
some sort of barebones program in windows (would appreciate help here as I really hate using the winapi)

csgo, with and without raw input, and with and without multicore
csgo with various fps capping mechanisms
cs 1.6
quake live. i have effectively no experience with quake so for those of you who do, please let me know which settings to mess around with

quake live. i have effectively no experience with quake so for those of you who do, please let me know which settings to mess around with
in_mouse (I use 2) and m_filter (I use 0)

http://www.quakelive.com/forum/showthre ... -1-1-and-2

http://www.quakelive.com/forum/showthre ... r-question

Re: my ghetto led lag-measuring setup

Posted: 19 Sep 2014, 19:41
by flood
first measurements :D

for a barebones vsync off love2d program on windows 8, i measured 3ms of input lag when slamming my dead g100s onto my new g100s. the program is running at 500fps and my g100s is 500hz and i'm using a crt.

repeating it now.

done:
from 9 more measurements:
5,4,5,3,5,3,3,4,5 ms

sample video:
https://www.youtube.com/watch?v=IA0T_o9qQVc

Re: my ghetto led lag-measuring setup

Posted: 19 Sep 2014, 20:14
by spacediver
crap, sorry flood!

I meant to reply to your post and ended up editing instead. argh!

Re: my ghetto led lag-measuring setup

Posted: 19 Sep 2014, 20:18
by flood
single measurement of mouse movement lag in csgo

opened a really simple custom map. no fps_cap, scoped into a region where the wall meets the sky, set sensitivity to 200, and hidehud 4.

all settings verylow/low, 4x msaa, multicore off.

the ingame fps was >900.

i count 5ms

video: https://www.youtube.com/watch?v=vlHEpju24-Q

Re: my ghetto led lag-measuring setup

Posted: 19 Sep 2014, 20:19
by flood
lol no worries i pressed back a few times and the original post was still in my browser's history

Re: my ghetto led lag-measuring setup

Posted: 19 Sep 2014, 20:23
by spacediver
whew ok, i tried the same with no luck :)

Re: flood's input lag measurements

Posted: 22 Sep 2014, 22:51
by flood
5 more csgo measurements: 6,6,5,5,5 ms

csgo measurements with fps_max 250: 7,6,6 ms