3S:OE Input Lag Testing

Tech Talk
1

I did a simple 60fps camcorder input lag test for 3S:OE. Result: both consoles have 4f input delay.

[S]This is identical to arcade ST and other Capcom fighters.[/S] I’m not positive if arcade 3S is the same, maybe someone else who knows can chime in.

[This is not identical to arcade ST as previously stated. Later in the thread the NKI video was posted and it was noted that NKI starts his count on the input frame, not on the next frame. Lately lag discussion in Tech Talk has revolved around frames of lag, counted as frames after the input frame. With that method, NKI has arcade ST at 3 frames.]

Tests were done on PS3 and 360, connected to a Sony GDM-FW900 CRT monitor via HDMI, through an HDFury DAC.

HDMI -> HDFury -> CRT

Resolutions tested were 480p (4:3), 480p (16:9), and 720p. Xbox360 was also tested at 1080p since it can upscale to that resolution. PS3 when configured at 1080p will still output 3SOE at 720p. No tests run for 480i, as I don’t have a setup for that right now.

Footage was recorded at HDV 1080i60 (shutter 1/60), imported with HDVSplit, processed with avisynth/virtualdub to separate the interlaced fields for analysis.

Input device was a Mad Catz Round 2 TE for Xbox360, modded with Toodles’ TE Kitty PCB for PS3 compatibility. This board lights up one of the 8 “turbo” LEDs on button actuation, corresponding to the pressed button, in both 360 and PS3 modes. The LED was used to time the 0 frame of input.

Hardware used

Spoiler

Console: “Slim” PS3
Console: “Fat” 360
Monitor: Sony GDM-FW900 CRT
DAC: HDFury original model HDMI/DVI -> VGA converter
Input: Mad Catz Round 2 TE for Xbox360
PCB: Stock 360 TE PCB
PCB: Toodles’ TE Kitty PCB for PS3 compatibility and LED activation
Camera: Canon HV20

Software used

Spoiler

Import: HDVSplit 0.77
Processing: AviSynth/VirtualDub (to separate interlaced fields for viewing)
Example AviSynth script:



DirectShowSource("3soe_lag_testing-0001.m2t")
complementparity
separatefields

Game settings

Spoiler

View: Normal
Filter: None
Scan Lines: Off
GGPO Delay: 0

Notes

Spoiler

I had some issues with VirtualDub occasionally dropping frames or ordering them improperly, so I made sure to get enough inputs to be certain of the result. Due to that I didn’t transcode any video samples, but will upload the original .m2t files imported from HDVSplit for anyone who wants to look at them. You’ll need something capable of splitting the interlaced frames into their separate fields; playback through a standard media player would only show 30fps. This could be avoided with a 60p camera but I don’t have one. :slight_smile:

Edit: Source files here: http://bitshare.com/?f=w7x0sxfp

Take note of the above or the files will not work for you. Also when analyzing the footage, keep in mind that the camera is not always in sync with the game/monitor, so sometimes the sequences will look like this: http://i.imgur.com/1kz1a.jpg

Tests were done in multiplayer local mode, Ryu vs Ryu in America/Alex stage. 1P st.jab as input, 50 times per test. Frame count starts at 0, when the LED lights, and ends at the first frame showing any portion of st.jab animation. Frame counts were consistent at 4 frames for each resolution on both consoles.

Sample GIFs linked below for each res/console. I’ll upload the original source content for people who want to poke at it (see Notes, above).

PS3:
480p 4:3 - 480p 16:9 - 720p

360:
480p 4:3 - 480p 16:9 - 720p - 1080p

What does this mean?

It means in the scenario outlined above, Ryu’s st.jab takes 4 frames to start animating after button input, on both consoles. It doesn’t mean there is no lag or weirdness or bugs anywhere else in the game. I won’t speculate on any of that. This is also on a CRT with a known lag-free analog conversion setup and good input devices. If your monitor lags, or your joystick PCB lags, or your cables are violating the laws of the universe, you may experience lag even if the game is working perfectly.

If anyone else can do similar testing, please post your results.

2

Anyone know how much lag 3S on the CPS3 board had?

3

Seeing you did this with the filters off if you did them with the filters on would that have caused lag or is that just simply the settings you chose?

4

Have you tested this in arcade mode? Scanline enable? Filter does affect you know.

5

Thanks for putting the Console differences B.S. to rest RoboKrikit, I knew it didn’t feel funny on the PS3!

Also, can you test to see if the filters will cause lag?

6

Well the test he did already had the filters turned off.

7

Alright great, thanks for testing… Now we just need to compare this to SFIII3rd on CPS3 and we’ll be all set. :slight_smile:

8

So, this is the same shit in both consoles ? Thanks you.

9

aaaaaaaaand how

10

“Shit” is yet to be determined since we haven’t tested how much input lag the original board had. As stated, older Capcom games (ST, Alpha 3, etc.) also had 4 frames of lag, but we’re not sure about 3rd Strike.

And speaking of “shit” anyone tries to start any of that here in my forum gets infracted, no questions asked. So think before you post. This is Tech Talk, we don’t allow the usual “shit” you see in the forums here.

11

Are we sure that st.jab shows animation on the first frame?
Might be better to measure until the first active frame, ‘startup’ frame data is readily available. It should be completely clear what frame the move hits.

Keep in mind that ‘startup’ is a misnomer, should be ‘hits on.’ A move with 3f ‘startup,’ actually has 2f startup, and hits on the 3rd frame.

12

Call me ignorant but is the HDFury proven to not lag when transcoding its video? Wouldn’t it have to buffer a frame into memory before it can output it?

13

Thank you for taking the time to do this. This goes a long way in helping end pointless arguments. Any chance we can get this tested against the arcade board?

Even testing an emulated version of 3rd strike on PC could prove helpful.

14

Emu’d on PC wouldn’t be much help as, depending on the emulator, the input delay there is different as well.

15

Irrelevant. The point is, the time from the moment the button is pressed until there is reaction on the screen is identical in both 360 and PS3 versions, all else being equal. What it proves is that no one console is slower/faster/laggier than the other.

Moot point since both systems went through it. The test isn’t meant to show that the HDFury is/isn’t more laggy than other options, but whether the Xbox is laggier than PS3 or visa versa. Since both systems went through it, any delay caused by the HDFury was present in both sets of tests.

Irrelevant to what he’s trying to show. His tests are for proving PS3 and Xbox360 latencies are the same. Whether filters affect the latency is a valid question, but separate from what he’s trying to show. Hopefully he can be convinced to test that question as well.

16

So… does this test debunk all of the “sky is falling” comments that everyone has been making re: PS3 being laggier than X360? Or is the jury still out pending verification of CPS3 3S 4f delay?

17

1.Yup. No difference between PS3 and Xbox360.
2. Latency difference between consoles and arcade versions has NOT been definitively answered yet, but until such time as it is, there’s good reason to believe the console versions match the arcade version; ST showed a similar 4f delay from button press to animation in the old arcade ST tests NKI did that damned CCC2. But it hasn’t been proven yet.

18

At this point, I’d have to wager to guess the majority of the complaints are coming from 3S purists who have never made the switch to next gen. As a result they probably aren’t too familiar with the tech behind HD setups/converters and what not.

I don’t really feel particularly comfortable saying people whom are way more familiar with 3S than I ever intend to be are wrong, but given that there seems to be a tendency to claim that the 360 version is better than the PS3 version, it certainly calls into question how accurate their sense of “feel” is on this one. I imagine we’ll see more conclusive info in the next few days.

19

I didn’t run the initial test with filters in the mix since it would’ve taken a lot more time. If I get some time I will do some filter comparisons.

The HDFury is very fast. I’m sure there is a way to measure the delay, but it is slight enough not to be significant when measuring in frames. FWIW, I have tested the native Xbox360 VGA cable vs Xbox360 HDMI -> HDFury, and they always tested the same for me. Like Toodles says, they are both going through the HDFury anyhow, so at least for 360 vs PS3 it doesn’t matter.

Keep in mind it is just a simple input lag test with no real gameplay going on. It won’t test slowdown or sound sync issues or any other game glitches. I don’t really play 3S so I won’t comment on whether or not either console’s game feels more like the arcade version or which is ‘best’.

Frame data like startup is counting from the start of the game initiating the move, not start of button press. If that were so then you’d have to have different frame data for every version of the game that has different input delay (i.e. SF4 360 vs PS3). The only frame data that could have different startup than the first frame of animation would be frame data direct from the game designers, since there would not be another way to measure when the move starts, unless it made a sound in advance or something.

It is typical for a move to begin animating on the first frame. Since they are animating at the same time they are getting the input at the same time. Testing the 3SOE frame data PS3 vs 360 is way out of scope, I’m not going there. :slight_smile:

20

Did you do this all in one pass? Because there is up to a 2f potential variation with this testing methodology.