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

GSYNC Input Lag Test Results in Chivalry: Medieval Warfare

Post by sharknice » 20 Jan 2014, 02:32

I did some input lag tests in Chivalry: Medieval Warfare similar to Mark's input test here.


Method:
The Corsair K95 keyboard has a mode built into the hardware where it lights a key as the switch is activated. I recorded the LED and game screen using a Nikon J1 high speed camera recording at 1200 fps.

I found the easiest thing to do an input test with was activating the voice macro menu because the menu appears instantly whereas with swinging a sword it is hard to tell when the animation begins swinging. I did do a few tests to make sure they both had the same input lag just in case they were different, but they had the same amount.

Using ffdshow's frame number display option and media player classic I counted the frames from when the LED was shown to when the menu appeared on screen. Then I multiplied the number of frames by .83333 ms (1000ms/1200fps) to get the delay in milliseconds.

For all of the tests I had my keyboard set to 1000hz polling and Chivalry frame capped to 120 fps using bSmoothFrameRate=true and MaxSmoothedFrameRate=120. The game would not go above 120 fps even with MaxSmoothedFrameRate set to higher values. The frame rate was at a constant 120 fps during the tests.

For the SLI tests I used Force Alternate Frame Rendering 2, mode 1 does not work right for Chivalry and only gets about 40 fps. GSYNC does not work in SLI, the game would crash or run at 6 fps.


Results:
Single GPU, GSYNC ON, Monitor@144hz
20.5 ms average

Single GPU, GSYNC ON, Monitor@120hz
28.5 ms average

Single GPU, Monitor@120hz
23 ms average

Single GPU, Monitor@120hz, ULMB ON
27 ms average

SLI, Monitor@120hz
26 ms average

SLI, Monitor@120hz, ULMB ON
30.5 ms average


Thoughts:
Similar to Mark's results GSYNC seemed to add some delay when frame rate was at the max refresh rate. Running at a constant 120fps GSYNC added an extra 8 ms of delay when the monitor was only set to 120hz versus 144hz.

Ultra Low Motion Blur (ULMB) added about 4 ms of lag.

SLI added about 3 ms of lag.

Since Chivalry will not run at more than 120fps I didn't try out 144hz @120fps without GSYNC because the tearing and stuttering would be so bad. For games that have a 120fps limit it seems GSYNC actually has an input lag advantage over NO SYNC on 144hz monitors.

User avatar
Chief Blur Buster
Site Admin
Posts: 11648
Joined: 05 Dec 2013, 15:44
Location: Toronto / Hamilton, Ontario, Canada
Contact:

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

Post by Chief Blur Buster » 20 Jan 2014, 03:16

sharknice wrote:Method:
The Corsair K95 keyboard has a mode built into the hardware where it lights a key as the switch is activated. I recorded the LED and game screen using a Nikon J1 high speed camera recording at 1200 fps.
That's pretty neat; if the keyboard lights the LED less than 1ms. I need to be sure, if this is, indeed, the case, and if so, then that's definitely a valid way of doing a high speed camera test. I'd love to see at least one of the high speed videos, if at all possible -- either the original MOV files or uploaded somewhere. I wonder if there's an online video service somewhere that allows you to frameskip backwards/forward signle frames..
sharknice wrote:Ultra Low Motion Blur (ULMB) added about 4 ms of lag.
This is the correct average difference, that is a good verification of the accuracy of the high speed camera test.
sharknice wrote:SLI added about 3 ms of lag.
Very interesting to know about this too.

Do you have a verification test of the LED accuracy, so we can be confident of its accuracy? (Though your ULMB test shows the accuracy, I'd like to also be sure, as well).
How many runs did you do to get the averages?

We'd love to publish an article about this if you're interested. I'd like to see tests of VSYNC OFF versus VSYNC ON versus GSYNC too, with at least a dozen runs.
Head of Blur Busters - BlurBusters.com | TestUFO.com | Follow @BlurBusters on Twitter

Image
Forum Rules wrote:  1. Rule #1: Be Nice. This is published forum rule #1. Even To Newbies & People You Disagree With!
  2. Please report rule violations If you see a post that violates forum rules, then report the post.
  3. ALWAYS respect indie testers here. See how indies are bootstrapping Blur Busters research!

User avatar
nimbulan
Posts: 323
Joined: 29 Dec 2013, 23:32
Location: Oregon

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

Post by nimbulan » 20 Jan 2014, 03:23

The results seem about as expected. An additional 1 frame delay when capped at the refresh rate and about 4ms of lag with ULMB on.

As an idea for further testing, there is a variable in Unreal Engine 3 that I've always been curious about: OneFrameThreadLag. This is set to true in every UE3 game I've checked and I'm curious if it affects input lag like it sounds like it will. It is found in UDKSystemSettings.ini.

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 » 20 Jan 2014, 03:57

Chief Blur Buster wrote:
sharknice wrote:Method:
The Corsair K95 keyboard has a mode built into the hardware where it lights a key as the switch is activated. I recorded the LED and game screen using a Nikon J1 high speed camera recording at 1200 fps.
That's pretty neat; if the keyboard lights the LED less than 1ms. I need to be sure, if this is, indeed, the case, and if so, then that's definitely a valid way of doing a high speed camera test. I'd love to see at least one of the high speed videos, if at all possible -- either the original MOV files or uploaded somewhere. I wonder if there's an online video service somewhere that allows you to frameskip backwards/forward signle frames..
sharknice wrote:Ultra Low Motion Blur (ULMB) added about 4 ms of lag.
This is the correct average difference, that is a good verification of the accuracy of the high speed camera test.
sharknice wrote:SLI added about 3 ms of lag.
Very interesting to know about this too.

Do you have a verification test of the LED accuracy, so we can be confident of its accuracy? (Though your ULMB test shows the accuracy, I'd like to also be sure, as well).
How many runs did you do to get the averages?

We'd love to publish an article about this if you're interested. I'd like to see tests of VSYNC OFF versus VSYNC ON versus GSYNC too, with at least a dozen runs.
I emailed you one of the mov files. I also emailed you one showing the numlock indicator versus the numlock button's LED. It seems to take around 25 ms for the indicator to switch after pressing the button.
You will notice the key LEDs actually use PWM, but it looks like the initial press starts with it fully on.

User avatar
RealNC
Site Admin
Posts: 3743
Joined: 24 Dec 2013, 18:32
Contact:

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

Post by RealNC » 20 Jan 2014, 07:01

You forgot to mention a crucial detail: Is this comparison against V-Sync enabled or disabled?
SteamGitHubStack Overflow
The views and opinions expressed in my posts are my own and do not necessarily reflect the official policy or position of Blur Busters.

User avatar
Chief Blur Buster
Site Admin
Posts: 11648
Joined: 05 Dec 2013, 15:44
Location: Toronto / Hamilton, Ontario, Canada
Contact:

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

Post by Chief Blur Buster » 20 Jan 2014, 10:28

RealNC wrote:You forgot to mention a crucial detail: Is this comparison against V-Sync enabled or disabled?
It's total-chain lag (button to pixels) so VSYNC OFF isn't a reference point, and thus isn't necessary to mention.

But yes, I am really curious of VSYNC OFF versus VSYNC ON versus GSYNC -- I need to do more tests of these too.
Head of Blur Busters - BlurBusters.com | TestUFO.com | Follow @BlurBusters on Twitter

Image
Forum Rules wrote:  1. Rule #1: Be Nice. This is published forum rule #1. Even To Newbies & People You Disagree With!
  2. Please report rule violations If you see a post that violates forum rules, then report the post.
  3. ALWAYS respect indie testers here. See how indies are bootstrapping Blur Busters research!

User avatar
Chief Blur Buster
Site Admin
Posts: 11648
Joined: 05 Dec 2013, 15:44
Location: Toronto / Hamilton, Ontario, Canada
Contact:

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

Post by Chief Blur Buster » 20 Jan 2014, 10:33

sharknice wrote:I emailed you one of the mov files. I also emailed you one showing the numlock indicator versus the numlock button's LED. It seems to take around 25 ms for the indicator to switch after pressing the button.
You will notice the key LEDs actually use PWM, but it looks like the initial press starts with it fully on.
Thanks!
I am curious though, if this is a CPU-driven interval. If so, then I might worry the 25ms might vary during the heat of the moment (heavy CPU usage). I have seen the latency of Numlock LED vary quite a lot based on how heavy the CPU is loaded. Can you start running a multithreaded bechmark and get all cores to 100%, and re-do the numlock high speed video test and see if the 25ms varies? To be safe, I need to see video of the numlock non-varying during both 0% CPU and 100% (all cores) CPU.

This does give me pressure to publish a mouse modding HOWTO, so that I can show how to make modifications.
Head of Blur Busters - BlurBusters.com | TestUFO.com | Follow @BlurBusters on Twitter

Image
Forum Rules wrote:  1. Rule #1: Be Nice. This is published forum rule #1. Even To Newbies & People You Disagree With!
  2. Please report rule violations If you see a post that violates forum rules, then report the post.
  3. ALWAYS respect indie testers here. See how indies are bootstrapping Blur Busters research!

User avatar
RealNC
Site Admin
Posts: 3743
Joined: 24 Dec 2013, 18:32
Contact:

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

Post by RealNC » 20 Jan 2014, 11:01

Chief Blur Buster wrote:
RealNC wrote:You forgot to mention a crucial detail: Is this comparison against V-Sync enabled or disabled?
It's total-chain lag (button to pixels) so VSYNC OFF isn't a reference point, and thus isn't necessary to mention.
Now I'm confused. With V-Sync on, there's no change in pixels anywhere on the screen until the next vblank, so "button to pixels" time can be affected severely.
SteamGitHubStack Overflow
The views and opinions expressed in my posts are my own and do not necessarily reflect the official policy or position of Blur Busters.

User avatar
Chief Blur Buster
Site Admin
Posts: 11648
Joined: 05 Dec 2013, 15:44
Location: Toronto / Hamilton, Ontario, Canada
Contact:

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

Post by Chief Blur Buster » 20 Jan 2014, 11:23

RealNC wrote:Now I'm confused. With V-Sync on, there's no change in pixels anywhere on the screen until the next vblank, so "button to pixels" time can be affected severely.
Your question was "Is this comparison against V-Sync enabled or disabled?".

Do you mean to ask: "You should add VSYNC ON tests too, for comparisions' sake?"
Or did you mean to ask: "Is the first two tests (GSYNC) compared against the last four tests as VSYNC OFF or VSYNC ON?"

If you meant the latter, good question. I automatically assumed sharknice tested VSYNC OFF (and I am pretty damn certain he did), but you are right, it wasn't implied. (sharknice, it's VSYNC OFF, correct?)
Head of Blur Busters - BlurBusters.com | TestUFO.com | Follow @BlurBusters on Twitter

Image
Forum Rules wrote:  1. Rule #1: Be Nice. This is published forum rule #1. Even To Newbies & People You Disagree With!
  2. Please report rule violations If you see a post that violates forum rules, then report the post.
  3. ALWAYS respect indie testers here. See how indies are bootstrapping Blur Busters research!

bbguy321
Posts: 3
Joined: 18 Jan 2014, 14:58

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

Post by bbguy321 » 20 Jan 2014, 11:51

Must be vsync off, otherwise baseline test results would have more lag instead of less.

Post Reply