1. Hey Guest, looking for Virtua Fighter 5: Ultimate Showdown content? Rest assured that the game is identical to Virtua Fighter 5: Final Showdown so all current resources on here such as Command Lists with frame data, Combo Lists and the Wiki still apply. However, you can expect some VF5US specific changes to come soon!
    Dismiss Notice

PS3 Vs. X360 - Why do PS3 FGs have input lag compared to X360 versions?

Discussion in 'Console' started by Junsuina Chikara, Mar 12, 2013.

  1. BlackGeneral

    BlackGeneral The Emperor of Elbows Bronze Supporter Content Manager Akira

    PSN:
    BlackGeneral
    XBL:
    BeeGee VFDC
    Oh... I have no idea. Old PS3 problems? Many people say that slowdowns (obvious ones like UMvC3) only happen on the fat PS3's. How true this is? No idea.
     
  2. Pai~Chun

    Pai~Chun Well-Known Member

    XBL:
    Lishao Tao GPK
    Round 2! :p

    /thread
     
  3. noodalls

    noodalls Well-Known Member Bronze Supporter

    Old topic I know, but I've been doing lots of work on input lag recently. I have created my own testing system, which I have been using on lots of different games. Details are here http://forums.shoryuken.com/discuss...es-controllers-etc-20-4-pc-games-added#latest

    Got around to testing VF today.

    https://twitter.com/noodalls/status/858884742300565506

    Basically, VF5 on PS3 (60.4ms), VF5FS on PS3 (60.3ms) and VF5FS on X360 (64.4ms) are almost exactly the same speed. VF5FS on X1 emulation (78.9ms) is about a frame slower. I don't have PS-Now here so no way to test that. Will probably get Yakuza 6 at some point if it's cheap enough in Japan.
     
    steelbaz, Kruza, Ellis and 1 other person like this.
  4. MarlyJay

    MarlyJay Moderator - 9K'ing for justice. Staff Member Gold Supporter

    PSN:
    MarlyJay
    XBL:
    MarlyJay
    I'd guess that the results for PS3 and 360 are within the margin of error of each other? Seems right, as they feel the same.
    I remember being told that the console versions of FS are slightly less responsive than the arcade version of the game. Wonder if that is actually true?
    Thanks for your work, Noodalls
     
    Ellis likes this.
  5. noodalls

    noodalls Well-Known Member Bronze Supporter


    The tests tend to vary by about 1ms between takes. The 4ms is a bit more than that. My theory (unproven) is that the X360 polls the controller less frequently than the PS3, so that even though the games are running at the same speed, the response to the controller will be a bit more spread out and on average a touch slower.
     
  6. MarlyJay

    MarlyJay Moderator - 9K'ing for justice. Staff Member Gold Supporter

    PSN:
    MarlyJay
    XBL:
    MarlyJay
    Different controllers have different lag values too, even on the same console. How do you normalise when testing?
     
  7. noodalls

    noodalls Well-Known Member Bronze Supporter


    I used the Brook UFB across the three different consoles. I haven't found a faster PCB than this yet.

    Edit - typo
     
  8. oneida

    oneida Long Arm of the Lau Silver Supporter

    PSN:
    oneida_vf
    XBL:
    oneida1
    Thanks for testing this, I had been meaning to do a formal test for a while but never got around to it.
     
  9. Myke

    Myke Administrator Staff Member Content Manager Kage

    PSN:
    Myke623
    XBL:
    Myke623
    I still maintain that there's a difference between the PS3 and X360 in terms of input delay, which I become acutely aware of whenever I drop Kage's [4][6][K][+][G][K]![K] or [6][2][3][P][+][K]![P]![K]![K] after having played one system for a prolonged period before switching over to the other. The results above show a difference of around 4 ms, which is about 1/4 of a frame so could one argue that an input might be "off" by an entire frame by the fourth input? I don't know, I feel like that's a crude approximation on my part.

    On a side note @noodalls, if you have the equipment, time (and can be bothered!) I'd really appreciate you explore the boundaries of the hit-timing requirements of the aforementioned Kage combos.

    Anyway, I watched your explanatory video but I struggled to follow your testing method. That aside, can you help explain what we're looking at with the report cards? Below is the X360 version using a Brooks Universal Fight Board (UFB):

    upload_2017-5-1_23-24-19.png

    I guess that the X-axis is the measured delay from input registered to display change, but what's the Y-axis represent? Also, what are the recorded values? Occurrence counts?

    Also, for the specific VF5FS tests, what was the actual input(s) used? I'm assuming you're aware of the 10f input buffer and avoided it skewing results?
     
  10. noodalls

    noodalls Well-Known Member Bronze Supporter

    I do commands every 30F to try to avoid overlapping commands. The move I used was Pai's standing [P]. I almost got caught with the change in speed from VF5 to VF5FS (i10 to i11).

    The numbers down the side tell you how late or early in the frame when the button is pressed. So when the button is pressed with 16.66ms left (i.e. right at the beginning of the frame) it will take 3F+16.66ms to occur (eight times). When the button is pressed at the half way point (8.66ms elapsed, 8ms remaining) 90% of the time it will take a further 3F+8ms (58ms) to respond and 10% of the time it will take (66.6 + 8) 74.8ms to respond. When the button is pressed in the final ms of the frame, it will always (10/10) take 1ms+4F = 67.6ms to respond.
     
  11. Myke

    Myke Administrator Staff Member Content Manager Kage

    PSN:
    Myke623
    XBL:
    Myke623
    How late or early in the frame? What frame? I was assuming that all timing began the moment an input was registered, but you're describing a relationship between when an input is registered and... something else. What is that something else?

    Do you have a wiring diagram or schematic that depicts the testing setup? I'm just trying to understand the round-trip of the signal from input to measured output.
     
  12. Myke

    Myke Administrator Staff Member Content Manager Kage

    PSN:
    Myke623
    XBL:
    Myke623
    I had trouble following your reading of the numbers, and the delays they equated to, so I tried to reconstruct your report card to expose some of the calculations which I think I've done correctly as verified by the same weighted average of 64.33 ms.

    upload_2017-5-2_13-36-38.png

    So essentially, you're calculating the input delay as:

    Time remaining when input registered + (No. of frames until output observed x 1/60s)

    I guess the thing I'm having trouble grasping, as described in my previous post, is this "time remaining" component.

    Also, I noticed there was 1 count of "Nil" in your report card for the X360 which I'm assuming is a dropped input. This 1 dropped input out of 176 inputs represents about 0.6% error, and as someone who uses a modded joystick and has countless times sworn that an input was dropped, I don't know how to feel about that. Is that considered normal? Are you able to do a test with a much larger sample size (say, a typical VF round's worth?) to see how many inputs are dropped? :D
     
    Kruza likes this.
  13. noodalls

    noodalls Well-Known Member Bronze Supporter


    https://twitter.com/noodalls/status/858884742300565506

    So,you are correct that Nil refers to a dropped input. For something like X360 VF5FS where only a single dropped input has occurred, I don't think this is anything to worry about. Further, it appears around the time of the change from 3F to 4F, suggesting that the command has fallen in between the two input windows.

    The way I send inputs its quite specific, lasting for exactly a frame. Given that the position at which inputs are read is not exactly the same each time, what has likely happened is that the input has come just a bit too late for the 3F window, but also ended a bit too early for the 4F window, hence it misses both. Realistically, if I held the inputs for longer than a frame, it probably would have just registered as a 4F input.

    It's worth mentioning, I don't include the missed commands in the calculation of the averages, so games/controllers that drop a lot of commands are probably calculated to be a bit faster than they actually are.

    Something like this might make it a bit clearer.

    https://twitter.com/noodalls/status/818239281122529281




    Now, if you look at the Zero delay controller for PS3.

    https://twitter.com/noodalls/status/845863324478717952

    Missed commands all over the place, 23/96 commands missed. This is definitely a problem. Interestingly, this controller had never been tested for input lag before, as lacking common ground makes it hard using standard methods.
     
    Kruza likes this.

Share This Page

  1. This site uses cookies to help personalise content, tailor your experience and to keep you logged in if you register.
    By continuing to use this site, you are consenting to our use of cookies.
    Dismiss Notice