Crazy, so the SF4 engine basically renders slightly more frames per second than gets displayed, independent of the video signal itself (and its fixed framerate)? So I guess the inputs get sampled at the end of each “internal” frame? I guess things aren’t so simple with modern consoles. This might be getting into derail territory (we’re talking SF4 now and not really adding to what 3SOE is doing), but in short I want to say thanks for educating me, and if you guys can point me to any resources (threads here on SRK, articles somewhere, whatever) dealing with how modern consoles/arcade hardware handles input (at the hardware and/or software level) it would be greatly appreciated as some late-night nerdy reading 
I’m not a techie guy concerning frame data, how many frames of this and that, but I work with post production, have the technical knowledge to work the footage and got curious about the count, so I humbly downloaded Robo’s sources, opened them in After Effects, deinterlaced and checked. It’s 4 frames from the moment the button is pressed until the first frame of the punch animation starts (including this first frame).
Button press [1] - [2] - [3] - First Frame of animation [4]
So I don’t know if you guys should count 3 frame lag, since the first frame (button pressed) was supposed to match the last (first of animation), and the animation actually starts 3 frames after button press.
Well, that’s my humble contribution.
Sweet, thanks for confirming the data. I know dick about post production, so yours and wuziq’s feedback has been helpful. I actually had something wrong with the processing of the files in avisynth, and got a better workflow from wuziq (this fixes the odd outliers I was getting, not the average, which is still 4f). I’ll update first post when I finish the current batch of stuff; getting data for filtered modes and the distribution tables for the original vids.
Despite NKI’s vid starting at 1f, people have been referring to lag lately with input on 0 and the count at wherever the resultant move begins. I hesitate to try changing that and cause any further confusion. In terms of lag difference though, if the expected ‘lag’ was 0, then it should be 3 frames lag. But I think of the baseline input for a game more as response time, which would be 4. So input lag is sort of a misnomer.
Oh, just to make clear, I didn’t check all the videos and actually made a quick test only. The method applied is correct though.
Thanks trapy for the great input!
Interesting. So I went back and reprocessed the videos using wuziq’s avisynth method, and verified that that output matched wiseless’ After Effects method, then checked every input and made a distribution list. The average is still 4, but I bet the results will surprise you.
This makes me wonder a lot of things. Among them: Is the PS3 version really faster, or is the Toodles PCB faster than the MCZ 360 TE PCB?
Right, and you can come up with little tests (I.e. human tests, like attempting a deep jump in combo) that can possibly give you a rough idea, even in fighting games.
However, this is different than what I’m talking about, where someone normally plays the game. People can’t normally play the game and tell what the input lag is within 2 frames. This has been proven over and over.
People think they can, and you end up with excuses based upon our own minds tricking us. People say “well I missed all my links, I never miss all of my links. Therefore, the PS3 is laggy.” That is what our brains tell us. Our brains look for patterns to explain things even when they don’t exist. It is human nature.
There is an excellent book on this topic: http://www.amazon.com/Fooled-Randomness-Hidden-Chance-Markets/dp/1587990717
edit- If you are playing on a CRT with everything properly hooked up, and
My friend did a test last yr comparing stock te pcbs to others, and found on average the madcatz te had about 1/4 frame more of lag compared to its fastest hori counterpart iirc.
Interesting. So, at 720p we have:
PS3 3.72f
Xbox360 3.76f
I’d say the difference is well within statistical anomaly range. Its interesting though, that on the 360 running at 480p 4:3 makes it worse (4.3f) than 720p. Not that any of that makes a practical difference, its still close enough to round to 4f.
Does this mean “crisp” setting lags?
mmm no… i just didnt try crisp yesterday 
but this is with crisp… i made around a dozen attempts and took photograms from one representing the group
[media=youtube]RuW2WvsYcsg[/media]
doesnt seem like capcom screwed up anything with those filters… but anyway, i keep the old graphics since I like it better
Have you tried not using the HDMI ports and using regular cable ports?
I remember someone even mentioning there was less input leg when they hooked the game up via VGA cable. Maybe the HDMI port is causing the lag.
The HDMI port on 360/PS3 does not cause lag. That was probably the HDMI input on the person’s TV. Some displays have different image processing for the VGA input, which can result in less total lag using VGA.
I got footage of all of the filter modes and a brief test of the VGA out on 360 just for comparison, but haven’t analyzed them yet.
I was just curious, I honestly don’t feel any input lag compared to GGPO and the DC version but I’m no expert at the game. Just reporting what I heard. Be nice to know for sure rather than rule it out though.
Hey Robo - can you run the test on an ASUs vh236h? Also what is the purpose of the hd fury?
If I test on the VH236H it will just introduce another level of unpredictable lag (VH236H lags normally around half a frame, spiking up to 2 frames sporadically). It is a good monitor for real-world play, but I’m using the CRT for the test for good reason.
The HDFury is a lagless DAC used to connect HDMI sources to displays that take VGA in. The HDFury is reliable; my previous tests between it and raw VGA/component were always the same, and it is frequently used in this sort of testing. If it helps assuade anyone’s discomfort, in Cicada’s 223BW/2333HD/VH236H side-by-side test, the monitor with the least lag (1.8ms weighted avg) was connected through an HDFury. And of course both consoles in these tests are connected through the same HDFury.
When I redo these, if I can get the component->BNC setup working for the PS3, I’ll be able to remove the HDFury. It shouldn’t change the difference between the consoles much, but it might shave some overall ms.
Aside from a CRT, what is the best monitor to buy to ensure the least input lag for 3s?
I ordered a VH236H online and will get it in a few days since I can’t stand the input lag from my optoma hd20 projector, but now I’m starting to get doubts whether I’ll still notice the 0.5-2f difference with the VH236H since I play very technical characters. I wonder if another setup with a superior LCD or through a HDFury would be better…
You’re not going to do much better than the Evo monitor when it comes to HD displays.
so does this game suck?