No announcement yet.

Measure your INPUT LATENCY

  • Filter
  • Time
  • Show
Clear All
new posts

    Measure your INPUT LATENCY

    Hello folks, I'm shooting some trouble, and I came across some strange numbers in my input latency. Here's how to measure it:

    Run some instant game against weak bot
    Open Console by pressing F10 (or your bind)
    Type: stat engine a little and.. Watch for INPUT LATENCY row...
    My latency jumps between 25-50 , which I consider too many in milliseconds (it might be measured in a different unit, though).

    What are you usual values? Where do you think this value stem from? DirectInput, engine, mouse driver?

    I found this while messing about with the demo (before the retail was released), on my system this value gets as high as 200ms (that's really noticeable even without the display). As for where the value comes from, I assume that it's the time between polling for data & processing the data from the input buffer, it goes up as load on the CPU increases.

    Just for reference, 50ms is approximately the same time as it takes to perform a mouse click (press & release). If you don't notice the latency when you're playing then it's not a problem. From the testing I did I found that mouse (input) lag becomes apparent when the lantency becomes greater than a few frame times, which makes sense when you think about it.


      Nice answer, thanks! I can surely notice 50ms lag, since I played 2k4 at pretty high level, where you learn to notice the difference between 20 and 60 ms ping. I guess you are right about the CPU load - mine CPU (Athlon x2 4800) is at full load all the time and I always get the feeling that my mouse movement is little ahead of the screen. Specially in close range fights. So it's probably CPU-introduced added latency.

      I guess higher end CPU could keep my latency at constant 20 ms, which should not be recognizable.


        It varies from about 20-60 for me, but mostly it hangs around 35. You might be able to shave a few milliseconds off that by increasing the USB polling rate (assuming you have a USB mouse of course). The default is 125Hz, which adds up to 8ms of latency; you can increase it to 1kHz, which adds no more than 1ms. Google should find a tool to let you do that in XP.

        In previous UTs there was an option to increase the mouse update rate, but I can't find one in UT3.


          I already tried the mouserate switcher. I have increased the polling rate to 500 Hz on my Logitech G1. The Latency of input remains still unchanged. Thats why I consider the theory about CPU-related latency more probable reason of this problem.


            What does this do exactly? Measures the response rate from your mouse in game?

            I know my mouse has a default refresh rate of 125hz in windows, so what is it that this "engine start" does if 125 would not be what it displays for me?


              I tested that too and I got some causality.

              In the map fearless I got a range of 16ms to 30ms (mostly around 18-22) and on deck I got a range of 24ms to 85ms (mostly around 32-40).
              First I thought it might depend on my fps, but as you can see on the screenshot I got total different latencys at the same place in deck while the fps diddn't change that much. Often the fps are low at the same time the latency increases, so I thought the source of the latencyvalue might also effect the fps.
              [ DM-Fearless while moving around ] [ DM-Fearless while fighting ]
              [ DM-Deck while standing ] [ DM-Deck while moving ] [ DM-Deck while spectating ]

              So the latency increases while moving/turning around in deck but not so much in Fearless. The FPS behaves very similar, but while playing I noticed that the FPS can drop without effecting the latency. That means, that it's not the actual FPS that has impact on the latency, but something that has an effect on both the values.
              In Deck more is rendered than in Fearless, so turning around means much more acces to the memory and therefore more data streaming through the bus and the CPU.
              This also explains that the latency is going down if you stand still for a sec.
              So in the end it depends on the map how badly your latency is influenced by movement i.e. during a fight.

              Everything under 30 feels ok for me, but above it is noticable... especially if the rendertime of a frame is high to and both are added.

              //edit: Oh and be aware of the fact that displaying the engine stats reduces you framerate! (from 80 to 68 in my case)

              greetz Stryke