GSYNC Input Lag Test Results in Chivalry: Medieval Warfare

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

Re: GSYNC Input Lag Test Results in Chivalry: Medieval Warfa

Post by flood » 25 Apr 2014, 20:58

glad to see that it's possible for gsync to work well just by capping the game framerate.

can you test the input lag with vsync off, no gsync, and no framerate cap?

User avatar
sharknice
Posts: 295
Joined: 23 Dec 2013, 17:16
Location: Minnesota
Contact:

Re: GSYNC Input Lag Test Results in Chivalry: Medieval Warfa

Post by sharknice » 25 Apr 2014, 21:30

Chief Blur Buster wrote:
RealNC wrote:I assume a PS/2 keyboard is needed for this though.
I did some tests in the past and found the keyboard LED latency is a bit variable. It's somewhat based on CPU workload, and a very slow/overloaded system (especially older verisons of Windows, like Window 95, running so many apps) can have a delay between pressing a CapsLock and the LED turning on. The keyboard driver software turns on the LED after receiving the keypress events. I feel it isn't trustworthy enough for mullisecond-accurate input lag tests, as it hasn't been constant-offset (at least in previous versions of Windows).
I also tested caps lock delay comparing the caps lock switch LED with the caps lock indicator LED and there was quite a bit of delay and it wasn't very consistent. I don't remember the exact amount but I think it was in the hundreds of milliseconds.

Trip
Posts: 157
Joined: 23 Apr 2014, 15:44

Re: GSYNC Input Lag Test Results in Chivalry: Medieval Warfa

Post by Trip » 26 Apr 2014, 07:31

sharknice wrote: I also tested caps lock delay comparing the caps lock switch LED with the caps lock indicator LED and there was quite a bit of delay and it wasn't very consistent. I don't remember the exact amount but I think it was in the hundreds of milliseconds.
Crazy that a simple state on state off key can create such a relativly huge amount of lag. I didnt think it would be that much and that inconsistent. I also own a mouse which responds on key presses by flashing an led even when the computer is shutdown (cooler master sentinel). Is this a correct way to test for latency or can it still add some considerable latency?

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

Re: GSYNC Input Lag Test Results in Chivalry: Medieval Warfa

Post by flood » 26 Apr 2014, 18:01

I tested scroll lock LED lag in windows 7

I got an average input lag of 34ms +- 4ms. (LED time - click sound time)

methodology:

keyboard is cm storm quickfire rapid with cherry mx blues

recorded video with an iphone 5 at 60fps. Each video frame covers 16.667ms (i.e. max possible amount of motion blur).
audio/video times were synced by using a binder clip as a clapperboard.

This is kind of hard to explain but I specifically focused on events that happened near the end of the frame with the event. This allowed precision much better than 16ms for video events

For audio, by looking at the waveform in audacity, it is easy to get time measurements with precision of 1ms since the cherry mx blue switches give a very sharp spike


edit: got pretty much the same result on linux.

in my case it could be just be related to the keyboard's mechanical design. i.e. lag between hearing sound vs electrical signal down wire

User avatar
sharknice
Posts: 295
Joined: 23 Dec 2013, 17:16
Location: Minnesota
Contact:

Re: GSYNC Input Lag Test Results in Chivalry: Medieval Warfa

Post by sharknice » 26 Apr 2014, 21:30

Trip wrote:
sharknice wrote: I also tested caps lock delay comparing the caps lock switch LED with the caps lock indicator LED and there was quite a bit of delay and it wasn't very consistent. I don't remember the exact amount but I think it was in the hundreds of milliseconds.
Crazy that a simple state on state off key can create such a relativly huge amount of lag. I didnt think it would be that much and that inconsistent. I also own a mouse which responds on key presses by flashing an led even when the computer is shutdown (cooler master sentinel). Is this a correct way to test for latency or can it still add some considerable latency?
It is most likely instant, you could probably find out by zooming in with a high speed camera and looking at the button as it is being pressed down and the light comes on.

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

Re: GSYNC Input Lag Test Results in Chivalry: Medieval Warfa

Post by flood » 26 Apr 2014, 22:40

sharknice wrote:
Trip wrote:
sharknice wrote: I also tested caps lock delay comparing the caps lock switch LED with the caps lock indicator LED and there was quite a bit of delay and it wasn't very consistent. I don't remember the exact amount but I think it was in the hundreds of milliseconds.
Crazy that a simple state on state off key can create such a relativly huge amount of lag. I didnt think it would be that much and that inconsistent. I also own a mouse which responds on key presses by flashing an led even when the computer is shutdown (cooler master sentinel). Is this a correct way to test for latency or can it still add some considerable latency?
It is most likely instant, you could probably find out by zooming in with a high speed camera and looking at the button as it is being pressed down and the light comes on.
@trip
your mouse's led should be instant.

the thing with caps/num/scroll lock is that your keyboard sends a message to the operating system, which then responds with the command to turn on the led.

I've never encountered hundreds of ms of lag in the lock button though. Except when I get bsods or kernel panics in linux and the button isn't responsive at all.

User avatar
sharknice
Posts: 295
Joined: 23 Dec 2013, 17:16
Location: Minnesota
Contact:

Re: GSYNC Input Lag Test Results in Chivalry: Medieval Warfa

Post by sharknice » 27 Apr 2014, 02:00

flood wrote:
sharknice wrote:
Trip wrote:
sharknice wrote: I also tested caps lock delay comparing the caps lock switch LED with the caps lock indicator LED and there was quite a bit of delay and it wasn't very consistent. I don't remember the exact amount but I think it was in the hundreds of milliseconds.
Crazy that a simple state on state off key can create such a relativly huge amount of lag. I didnt think it would be that much and that inconsistent. I also own a mouse which responds on key presses by flashing an led even when the computer is shutdown (cooler master sentinel). Is this a correct way to test for latency or can it still add some considerable latency?
It is most likely instant, you could probably find out by zooming in with a high speed camera and looking at the button as it is being pressed down and the light comes on.
@trip
your mouse's led should be instant.

the thing with caps/num/scroll lock is that your keyboard sends a message to the operating system, which then responds with the command to turn on the led.

I've never encountered hundreds of ms of lag in the lock button though. Except when I get bsods or kernel panics in linux and the button isn't responsive at all.
It might have been 30 when I tested it, I never wrote it down or anything, I just remember it was significant.

Trip
Posts: 157
Joined: 23 Apr 2014, 15:44

Re: GSYNC Input Lag Test Results in Chivalry: Medieval Warfa

Post by Trip » 27 Apr 2014, 06:45

flood wrote:I tested scroll lock LED lag in windows 7

I got an average input lag of 34ms +- 4ms. (LED time - click sound time)
8 ms of variability in latency is indeed quite significant though maybe an iphone is just not fast enough since the interval is 16.6667 ms in which all kinds of stuff could happen. But indeed no further testing is nessecary I think the key is just too inconsistent especially when used in a video game which puts more strain on the system and maybe slows down response even more.
flood wrote: @trip
your mouse's led should be instant.
Yeah I already thought it would be if the computer is in the off state and the led still works it must be a local circuit in the mouse. Which should add no delay at all. Shame the lock keys dont work this fast since it would be great for everyone to have an easy way of testing input lag.

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

Re: GSYNC Input Lag Test Results in Chivalry: Medieval Warfa

Post by flood » 27 Apr 2014, 18:06

I didnt mention but +- 4ms is a 95% confidence interval.
the standard dev was like 2.2ms or something

yes that's far lower than 17ms, and that's because I cherry-picked frames that look very similar (not frames with similar timings). I only measured the time deltas after selecting the frames, so my data isn't biased to have low variance.

anyway it may be reasonably consistent in my test, but caps lock isn't a good idea because we can't be sure that it will perform as well when the cpu is loaded by a game. a local circuit + led is best for measuring visual input lag

Post Reply