Tested Input Latency of TGM1-2-3, Tetris Friends, Playstation Sakura.

Thread in 'Research & Development' started by JonnyFacepunch, 20 Aug 2015.

  1. I was curious to test the actual input latency of the different Tetris variants I've been playing on my (on paper) suboptimal system.

    Other threads report that both Windows 7 64 bit and the 'keyboard input hack' are dealbreakers for running TGM3 and even WinXP and a standard PC CRT will be more than 8 frames of input latency, but I measured on the low end of arcade clone WinXP+Arcade IO+lagless display results with only a Win7x64 system+LCD monitor in TGM3.

    Results:
    Actual time from keypress to piece moving on monitor (keyboard->indicator latency)+(indicator->display latency):

    Tetris The Grandmaster shmupmametgm: 5 frames
    TGM2+ shmupmametgm: 4 frames
    TGM3: TI default keyboard input hack: 6 frames
    Tetris with Cardcaptor Sakura Eternal Heart epsxe windowed: 6 frames
    Tetris Friends High Quality /w Adblock, Chrome: 6 Frames
    edit
    Texmaster 3 frames
    There was no difference between windowed and fullscreen latency for any TGM game on my system.

    I used a novel testing method that I don't believe has been used elsewhere but is easily done on any system by anyone who has a 240fps or better camera.

    Method:
    1.Use 240fps camera to determine latency between capslock keypress and capslock indicator (measured as 4frames@240fps, equivalent to 1frame@60fps on my keyboard, consistently reproduced).

    2.Rebind left input to capslock key in game. Use 240fps camera to determine latency between keyboard capslock indicator and piece moving left on screen. Add both to determine actual latency.

    My PC:
    Win7x64
    Geforce GTX650Ti
    i5-2400 12GB RAM
    BenQ G2222HDL DVI

    Using a CRT monitor would be ideal and further reduce these latencies (presumably significantly, perhaps a frame or more). I will try to get access to one.

    I was just counting frames manually advancing one by one and doing my math in a notepad, but I believe this method is quite accurate. Am I just looking at old information and this is normal performance, or does this appear to be much better than expected for my configuration? Is my method horribly flawed? Putting together an analysis/proof video would be an lot of work if this is not contrary to what is currently believed.
     
    Last edited: 25 Aug 2015
    colour_thief likes this.
  2. Wow, I wouldn't have expected so much for Sakura. I thought the Playstation was emulated really well.
     
  3. 6 frames is not much, because apparently the LCD monitor JonnyFacepunch has already 4 frames (about 65 or 70 ms) of response time. So 2 frames is actually really good for an emulator, I’m playing on PCSX2 and it’s like a quarter-second, it’s so heavy it’s almost unplayable lol

    Nice job measuring the input latency, I always knew it wasn’t THAT bad with TGM3, and really it’s not =p although yeah I already knew TGM2 would win (it’s like impossible to feel any lag on my CRT setup)
     
  4. Muf

    Muf

    Yeah, try again with a CRT.
     
  5. Altimor

    Altimor a.k.a. Ghett0

    Where's 65-70 ms from?
     
  6. Zaphod77

    Zaphod77 Resident Misinformer

    1000 ms in a second

    60 frames in a second
    1000/60=16 and 2/3 ms in one frame

    times 4 = 66 and 2/3 ms.

    simple math.:)
     
  7. I'm curious what Texmaster gets.
     
  8. Testing in MAME by pausing the emulation, holding a direction and then advancing one frame at a time until the piece moves:

    tgmj: 3 frames
    tgm2p: 2 frames

    Of course that ignores input device, OS and monitor lag etc.
     
  9. Altimor

    Altimor a.k.a. Ghett0

    Is this shmupmame or regular MAME?
     
  10. Regular MAME 0.164. I don't have shmupmame, but if you try it in shmupmame please post your results.

    Also, to clarify my previous post: By 2 frames I mean I have to advance twice until the piece moves. Maybe you could call this 1 frame of input lag, since the piece doesn't move for 1 frame.
     
  11. CPN

    CPN

    Must not be Shmupmame. TAP in Shmupmame has no frames of lag.

    MameTGM has one less frame of lag in TGM than other versions of MAME. Ideally you'd use Shmupmame for TAP and MameTGM for TGM1.

    Edit: there you go
     
  12. Altimor

    Altimor a.k.a. Ghett0

    TAP shmupmame still seems to be wonky, I have problems with quick up+downs not registering when I can do it 100% of the time on Tex/Ti.
     
  13. CPN

    CPN

    No problems in like 150 hours logged on 4.2 GUI.
     
  14. Altimor

    Altimor a.k.a. Ghett0

    Are you on stick or keyboard? It might have something to do with how it handles up+down on the same frame.
     
  15. CPN

    CPN

    Stick. I assume you're using keyboard...?
     
  16. Altimor

    Altimor a.k.a. Ghett0

    Yeah. It uses the last pressed key which is the same as altinput and I think Tex, so I have no idea what the problem is.
     
  17. Just figured out the config and made a texmaster vid. 3 frames. Texmaster is definitely more responsive than anything else.

    I haven't been able to find input latency for my specific monitor online. Is 65-70ms verified from some source?

    Still trying to get my hands on a CRT monitor...
     
  18. Only MAMETGM or ShmupMameTGM are last direction pressed, since upy wrote that patch to MAME's 4-way behavior. The default behavior in mainline MAME (and hence, ShmupMAME or any other fork that didn't pull in the MAMETGM mod) is that opposing directions cancel.
     
  19. Altimor

    Altimor a.k.a. Ghett0

    shmupmametgm is what I have. Forgot to mention that.
     
  20. Is it possible to use Mouse Script and whatever method that you guys mentioned here on the genuine TGM3 on Taito Type X machine? Would that make reduce input latency?
     

Share This Page