Super Turbo Offline Setup Guide & Lag Rating

2»

Comments

  • RufusRufus An unexpected database error has occurred. Joined: Posts: 1,966
    ...
    This test shows arcade SF2 having 3 frames of lag. The video calls it 4 frames because he counts them wrong. (sorry nki. according to nki's counting methods, if the game responds to your button press on the exact same frame that you made the press with no delay whatsoever, using perhaps some form of time travel, that counts as a delay of "one". that just ain't right)
    ...
    In NKI's defense, the video is about version differences, and he counts the frames for everything the same way.

    Also, I was looking in the wrong place. Here're lag testing images from CPS-2.

    http://www.pedantic.org/~nate/SF2T/lag/turbo3/
    Hitboxes http://www.pedantic.org/~nate/HDR/
    "You don't know what you're talking about as much as I do." -- Unknown
  • geadomgeadom Reversal Counter Hit Joined: Posts: 1,085
    Nice thread! Any chance to also post the Ps2 versions delays? (HSF2 and CCC2)
    "He´s different, he´s special." Pick up your spear, the only thing you can do is your best
  • Hyphen-atedHyphen-ated Joined: Posts: 39
    Rufus wrote: »
    Also, I was looking in the wrong place. Here're lag testing images from CPS-2.

    http://www.pedantic.org/~nate/SF2T/lag/turbo3/
    Great. This shows 3 frames of delay. You originally posted 2.5; does that mean you got only 2 frames of delay on some runs?
    I downloaded NKI's 100MB video and it shows 3 frames several times and 4 frames once.
    Maybe he's getting an additional frame of lag from his video converter. That would explain the discrepancy of 2-3 in your tests and 3-4 in his tests, because your measurement method bypasses video converter lag.

  • RufusRufus An unexpected database error has occurred. Joined: Posts: 1,966
    Great. This shows 3 frames of delay. You originally posted 2.5; does that mean you got only 2 frames of delay on some runs?
    It's been a while, but IIRC, I would have called that 2.8 frames:
    0.8 Frames for frame 0
    1.0 Frames for frame 1.
    1.0 Frames for frame 2.
    0.0 Frames for frame 3.

    I don't see particular clarity on how much lag to assign for frame 3 - do I count it as lag or not? Using 'when it actually shows on the screen' means that there's 0.6 frames before the game gets to Ken's sprite, but that would also mean that jumping attacks have less lag, and low attacks have more lag since the screen updates from top to bottom.
    I downloaded NKI's 100MB video and it shows 3 frames several times and 4 frames once.
    Maybe he's getting an additional frame of lag from his video converter. That would explain the discrepancy of 2-3 in your tests and 3-4 in his tests, because your measurement method bypasses video converter lag.
    I think that NKI is more careful than that. The discrepancy could easily be explained by different counting methods (maybe 'frame 3' should be counted as lag) or the limited time resolution of a 60FPS camera.
    Hitboxes http://www.pedantic.org/~nate/HDR/
    "You don't know what you're talking about as much as I do." -- Unknown
  • Hyphen-atedHyphen-ated Joined: Posts: 39
    Ohhh I see, you're using the height on the screen of the visual change to determine fractions of a frame.

    If you're going to do it that way I think you definitely don't add anything for frame 3. The "when it actually shows on the screen" definition is kind of assuming a granularity of one frame at best.

    Trying to measure below one frame runs into complications like the fact that the input sampling frames and the display frames don't happen at the same time and the difference between them is kind of hard to measure.
    Rufus wrote: »
    I think that NKI is more careful than that. The discrepancy could easily be explained by different counting methods (maybe 'frame 3' should be counted as lag) or the limited time resolution of a 60FPS camera.
    Yeah, I think the only discrepancy I was seeing is that your result of 2.5 gets counted as "3" when you don't have sub-frame measurement. If your test runs included any values above 3, those would be equivalent to the "4" that NKI's test sometimes produced (as I counted it)
    And the chart in the OP is additionally using NKI's counting method where he starts at 1 instead of 0. So your testing, NKI's testing, and the chart are all in agreement as far as CPS2 is concerned.

    Taking these things into account, there's apparently an actual discrepancy of 1 frame on the PS3 and 360 between your testing and the chart. That's small enough that I don't really care. Okay cool.

    I'm going to borrow my roommate's 60fps camera and test my laptop shmupmame setup that I had some people playing on at evo.
  • Hyphen-atedHyphen-ated Joined: Posts: 39
    edited August 2013
    I ran some tests on my setups for fun. These numbers are all based on counting from 0 (the correct way, hoho) so they're 1 lower than what the chart in the OP would say. Other than that they're pretty much in agreement with the chart, assuming my 22" monitor has around 1 frame of lag.

    ggpofba on my 22" lcd: 5-6 frames
    ggpofba on my 30" korean ips: 7-8 frames
    ggpo online on my 30" korean ips: 8-9 frames (i play on smoothing 1 so this is what I expected)
    (not measured but inferred: ggpo online on my 22" monitor: 6-7 frames. this is what 99% of my ST time has been on)
    shmupmame on my 22" monitor: 4-5 frames
    shmupmame on my laptop: 4-5 frames
    shmupmame on my 22" monitor with ps2 keyboard instead of usb arcade stick: 5-6 frames

    My test setup was this: as p1, push p2 into the corner, then throw by hammering on the hp button as fast as i can while filming with a 60fps camera. No LED, sadly, but it was usually pretty clear what frame the button was pushed on. I ran each test 10 times and I always got either the low number or the high number.

    It's nice to know exactly how much laggier my 30" is than my 22"

    I'd heard that keyboards are laggier. I didn't know whether to believe it. Now I know: for me the keyboard was about one frame laggier.

    I'm happy that shmupmame on my laptop is only about 1 frame worse than arcade. I could tell it was a lot closer to arcade than ggpo is. After playing a bunch of casuals on it at evo it improved my game on the superguns a lot. Too bad that was after the tournament was over.
  • RufusRufus An unexpected database error has occurred. Joined: Posts: 1,966
    ...

    I'd heard that keyboards are laggier. I didn't know whether to believe it. Now I know: for me the keyboard was about one frame laggier.

    I'm happy that shmupmame on my laptop is only about 1 frame worse than arcade. I could tell it was a lot closer to arcade than ggpo is. After playing a bunch of casuals on it at evo it improved my game on the superguns a lot. Too bad that was after the tournament was over.

    You might see better results with a dedicated gaming keyboard, custom drivers, or reworking the polling rate.
    Hitboxes http://www.pedantic.org/~nate/HDR/
    "You don't know what you're talking about as much as I do." -- Unknown
  • Hyphen-atedHyphen-ated Joined: Posts: 39
    edited August 2013
    I was still curious about the input delay of various non-SF2-related things so I ran some more tests. I'm posting them here because, hey, why not.

    The first 3 of these were done with my usb fightstick + joytokey. The PCB in it is from a madcatz 360 pad. The rest of the tests were with my ps2 keyboard. All were on my 22" lcd which I'm pretty sure adds one frame of lag.
    • (subtract 1 from these to get delay of the program itself not counting LCD lag)
    • Entering text in notepad with stick: 3-4f
    • Unity3d test program with stick: 2-3f
    • Pygame test program with stick: 3-4f
    • (subtract 2 from these to get delay of the program itself not counting LCD and keyboard lag)
    • Entering text in notepad with keyboard: 4-5f
    • Unity3d test program with keyboard: 3-4f
    • Pygame test program with keyboard: 4-5
    • Shooting in Dystopia with keyboard: 3-4f
    • Shooting in Science and Industry with keyboard: 3-4f
    • Shooting in The Binding of Isaac with keyboard: 5-6f
    • Drawing to the screen with js canvas operations in chrome (triggered by keyboard): 5-6f

    Unity3d is a game programming framework. I made a very simple program in it where pressing a key teleports an object to a new location.
    Pygame is a different game programming framework. I made a very simple program in it where pressing a key draws a bitmap to the screen.

    Dystopia is a Source engine (Half-Life 2) mod that I played and also coded on for a long time. http://www.dystopia-game.com/
    Science and Industry is a Half-Life 1 mod that I play: http://scienceandindustrygame.com/
    The Binding of Isaac is a flash game that I play

    Conclusions: That's awesome: Unity, HL1 games and HL2 games have only around 1-2 frames of lag in the game itself. (assuming my subtracting 2 because of the LCD and the keyboard is correct, which i'm pretty sure it is.) Take that moonchilde ;-)

    At one point I was slightly worried that joytokey would add latency. It's apparent that it doesn't, or at most it is by a very small amount

    Tests I'm going to do later: SDL and the Allegro game programming library.
    Rufus wrote: »
    You might see better results with a dedicated gaming keyboard, custom drivers, or reworking the polling rate.
    Do you know of any such custom drivers? I tried googling around and didn't find much.
    Also, PS2 uses interrupts, not polling, yeah? Unless you mean speeding up the clock on the PS2 bus? The one that runs at 10kHz at minimum

    Post edited by Hyphen-ated on
  • RufusRufus An unexpected database error has occurred. Joined: Posts: 1,966
    edited August 2013
    ...
    Do you know of any such custom drivers? I tried googling around and didn't find much.
    Also, PS2 uses interrupts, not polling, yeah? Unless you mean speeding up the clock on the PS2 bus? The one that runs at 10kHz at minimum
    I was thinking USB, but special drivers are not something I've really looked into. Technically, HID limits keyboards to 6 keys + 4 modifiers, but it looks like NKRO keyboards fake it by being multiple endpoints instead of using custom drivers.

    For AT and PS2 keyboards, the device generates the clock, so speeding it up might prove quite tricky.

    Hitboxes http://www.pedantic.org/~nate/HDR/
    "You don't know what you're talking about as much as I do." -- Unknown
  • jdubsjdubs Joined: Posts: 139
    You might see better results with a dedicated gaming keyboard, custom drivers, or reworking the polling rate.
    Do you know of any such custom drivers? I tried googling around and didn't find much.
    Also, PS2 uses interrupts, not polling, yeah? Unless you mean speeding up the clock on the PS2 bus? The one that runs at 10kHz at minimum

    [/quote]

    What MB are you using? If it is relatively modern AND has a PS/2 input, there's a good chance the PS/2 input is tied to a PS/2 to USB converter.

    -Jim

  • Hyphen-atedHyphen-ated Joined: Posts: 39
    jdubs wrote: »
    What MB are you using? If it is relatively modern AND has a PS/2 input, there's a good chance the PS/2 input is tied to a PS/2 to USB converter.
    Oh interesting, I never thought of that. I have an ASRock B85 pro4; it's pretty modern.
    My keyboard shows up in the windows device manager as "Standard PS/2 Keyboard" and the driver it's using is i8042prt.sys, which is a PS/2 driver. It also has its own IRQ assigned. These things lead me to believe that my motherboard is not using a PS/2 to USB converter. Sounds right to you?
  • jdubsjdubs Joined: Posts: 139
    Sounds like you're good to me!
  • undamnedundamned Wake up! Time to die! Joined: Posts: 1,686
    For those of you curious about upscaler setups, I just did an example setup for kicks:

    http://forums.shoryuken.com/discussion/comment/8368543/#Comment_8368543

    -ud
2»
Sign In or Register to comment.