Testing CCC2 For Input Lag

[NKI]

If you haven't been keeping up with Sirlin's thread "CCC2 info" (http://forums.shoryuken.com/showthread.php?t=141886), I'll briefly summarize why this video was necessary...

After Evo2007, I was quite vocal about my distaste for CCC2 because of its input lag. Several people claimed that CCC2 is just fine, and that I was "making up" this input lag as an excuse for why Gian didn't win at Evo. We needed some solid evidence one way or the other, so I took the time to put this vid together. (And it did take quite a bit of time.)

http://nki.combovideos.com (It's the last one listed under "Miscellaneous vids".)

Hopefully it's self-explanatory, but I'll be happy to answer any questions you may have.

And if anyone cares, the song used is Ghostwriter (remix) by RJD2.

[KayinNasaki]

The thing thats kind of freaky here is DC sometimes lagging. I hope that isn't an issue for HD Remix -- though I'm sure you've already communicated this to Sirlin.

[NKI]

Well, I am very hesitant to say this, because I really want the DC version's input lag to get fixed in STHD, but...it's not a huge deal. Most people (myself included) probably can't detect one frame of input lag just by playing the game, especially when the lag isn't constant. (Remember, sometimes the DC version has no lag.)

[DoughBoy]

I'm sure there will be plenty of people who will jump on in support or deriding the relevance of your test. So I'll just go ahead and thank you for taking the time and effort to produce this vid. As someone who only recently took up ST, I'm glad that there are people who care enough about the scene to do this.

[Havoc]

Well, science is science.

Method seems sound. Good job.

[shoo]

wow crazy, didnt know it was that bad

D:

good song too

[Keits]

Thanks for putting the time into this NKI. You are doing science and I'm, still alive.

edit - wow, timing it to the music! maybe you put too much time into it! but it was an enjoyable watch that way.

[goodm0urning]

Very illuminating. Thanks, NKI.

[aiy1tm]

Mmm, the delicious taste of science. So refreshing after spending too much time listening to people in the 9/11 CT thread.

Congratulations.

[Demon Dash]

Cool vid!

[JAMMAR]

Hahaha poor Sirlin. Vid was good. I knew there was something odd with my O. Sagat tiger shots when I first booted up CCC2 for the first time... I just thought it was just me.

[Toodles]

Good shit NKI

[Onslaught2000]

It's still so sad that this version turned out to be the worst yet. :sad:

I guess just have to wait for STHD now.

[Silks]

It's still so sad that this version turned out to be the worst yet. :sad:

I guess just have to wait for STHD now.

Worst? Wasn't the 3do version pretty terrible? :wink:

[Pablo_the_Mex]

Thanks. Maybe they can learn from this when making STHD.

[xqoole]

Quick question, do all psx to dc converters lag at least 1 frame? Specifically, how does the Innovation psx to dc one do?

[HP87]

Really nice vid. But why does the arcade have 4 frames of imput lag? That's just weird. I've always thought arcade has none. When tested on an emulator(fba), there should be only one frame of imput lag for ST due to the sound. The arcade lags more than emulator? o_O

[Monte]

very nice work NKI. glad you were able to prove people wrong especially when they take cheap personal shots at you.

[shoultzula]

damn, thats one hell of a way to test something.

and 8 frames of input lag should be noticeable to any player that plays fast frame games like ST and mvc2.

is it possible to stop kicking this dead horse now?

[Humbag]

Which version of Ghost Writer is that? Sounds different than the one i have.

[NKI]

Quick question, do all psx to dc converters lag at least 1 frame? Specifically, how does the Innovation psx to dc one do?I can't say for sure that all converters do, but I tried two totally different brands (one was Innovation, the other was Total Control 2), and they both had 1 frame of additional lag.

The arcade lags more than emulator? o_OWhen I fire up Kawaks and frame-step, it takes 4 frames for my move to come out, just like in the arcade version.

I dunno what's going on with the other emulators, but maybe it has something to do with modern PC's being vastly more powerful than CPS2 hardware.
Maybe a CS wizard like Toodles or Mike Z would know more about it...?

glad you were able to prove people wrongThis wasn't so much about proving people wrong. It was all about getting correct info out there and backing it up. Way too often people make claims, and when you call 'em out, they've got nothing to back it up with. I don't want to be one of those people.

and 8 frames of input lag should be noticeable to any player that plays fast frame games like ST and mvc2.To clarify, CCC2 only has 4 frames of lag compared to the arcade version. (The arcade version takes 4 frames to respond, CCC2 takes 8 frames to respond, so that's a 4-frame difference.)

Which version of Ghost Writer is that? Sounds different than the one i have.It's track #2 on The Horror maxi single.

[SweetJohnnyV]

Really cool vid NKI :tup:

[Silks]

I don't wanna be a party pooper, but could you have done the vid at 60 fps instead of 30?

Other than that, nicely done.

edit: to be more specific, maybe a 60 fps version of the video without the slow motion. Just the raw, real time capture. I want to see them side by side in real time is all

[gridman]

Thank God for NKI. I always wondered why I could tick perfect on arcade versions but not on CCC2. I told people (in person) that it lagged, got bashed, blah blah blah. Good shit man

Also - why no AE?

[Keits]

damn, thats one hell of a way to test something.

and 8 frames of input lag should be noticeable to any player that plays fast frame games like ST and mvc2.

is it possible to stop kicking this dead horse now?

Lets be realistic, and fair, here. If the arcade version's c.hk comes out in 4 frames, then this becomes the base for additional 'delay' in other versions.

So if arcade ST does it in 4 frames, and CCC2 does it in 8, then ccc2 has 4 frames of delay, not 8.


Did I word this correctly? :sweat:

[Humbag]

I assume he did for the frame tests?

Yeah, test AE as well just for shits and giggles.

[JubeiNinja69]

i like the test. someone else can also test it for double confirmation. now there was another issue about DC ST hit boxes. maybe a test can be done whether the hit boxes are correct or not.

[Josh-TheFunkDOC]

NKI, I can't say "thank you" enough! I hugely appreciate the effort you've put into this. Only one issue:

When I fire up Kawaks and frame-step, it takes 4 frames for my move to come out, just like in the arcade version.

OK, I think there might be some error with this.

When you say this, do you mean that you pressed the frame-advance key 4 times?

If that's the case, then it would be 2 frames of lag. All games read an input on the frame after you initiate it, so with no lag a move will come out after frame-advancing twice.

Final Burn Alpha, without sound, has no lag for Super Turbo. Kawaks has 2 frames. Subtract the first two times you frame advance to get the accurate number.

So would this actually be 6 frames for CCC2, etc.? Even that's worse than Kaillera P2P on Kawaks...

Thanks again,
Josh.

[guilers]

It's still so sad that this version turned out to be the worst yet. :sad:

I guess just have to wait for STHD now.

I remember when this title was first released and I got flamed for saying it was total garbage way to slow music problems frames skipping in certain stages.I guess it wasn't just my imagination.

[guilers]

told you guys ps1 version was the closest to the arcade version but even so this game guile always slow down when doing combos which throws of my timing.

[guilers]

damn, thats one hell of a way to test something.

and 8 frames of input lag should be noticeable to any player that plays fast frame games like ST and mvc2.

is it possible to stop kicking this dead horse now?

just goes to show the folks who were saying thi version of st was the GREATEST had no idea what they were talking about imho.

[Kinniku]

Great video.

[xqoole]

Well, I am very hesitant to say this, because I really want the DC version's input lag to get fixed in STHD, but...it's not a huge deal. Most people (myself included) probably can't detect one frame of input lag just by playing the game, especially when the lag isn't constant. (Remember, sometimes the DC version has no lag.)

Just to clarify for the DC test, you didn't use the triggers or shoulder buttons on the official dc padhack for the vid right??

[NKI]

why no AE?Simply because AE is not ST. (And because I hate AE.)

could you have done the vid at 60 fps instead of 30?All the original footage was shot at 60 FPS. Once I slowed it down frame-by-frame, there was no need to use 60 FPS for the final product.

But if you want to see the original 60 FPS videos, I've uploaded them all:
http://nki.combovideos.com (They're in a RAR file next to "Testing CCC2 For Input Lag".)

When you say this, do you mean that you pressed the frame-advance key 4 times?

If that's the case, then it would be 2 frames of lag. All games read an input on the frame after you initiate it, so with no lag a move will come out after frame-advancing twice.

Final Burn Alpha, without sound, has no lag for Super Turbo. Kawaks has 2 frames. Subtract the first two times you frame advance to get the accurate number.Ah, I see. Well, I'm not sure what's going on with the emulators, but obviously they do not behave the same way as the actual arcade board. If I cared enough, I could test out the input lag on emulators running in real-time (which I think is much more relevant than emulators stepping frame-by-frame), but unless Evo decides to start using Kawaks for Evo, I really can't be bothered to wire up all that stuff again.

So would this actually be 6 frames for CCC2, etc.?No, because when you watch the video of the two running side-by-side sharing an input via the Y-splitter, CCC2 lags 4 frames behind.

Just to clarify for the DC test, you didn't use the triggers or shoulder buttons on the official dc padhack for the vid right??Correct. I used the 'Y' button on a standard Sega brand controller.

[UltraDavid]

Most surprising thing to me here is that the arcade version doesn't respond for four frames. What's the deal with that? Shouldn't it respond on the next frame?

[BoggleMinds]

Most surprising thing to me here is that the arcade version doesn't respond for four frames. What's the deal with that? Shouldn't it respond on the next frame?

For the arcade version, inputs register on the next frame, and for some reason it then takes two more frames for the move to begin.

[Shodokan123]

Analog trigger buttons (on the official dc controllers) are proven to have a 1 frame delay, so that translates directly to the game's extra added frame from ryu's c.hk. If using psx>dc converter there should be no 1 frame delay... i beleive that the converter outputs the signal as digital and not analog. I could be wrong though and i think it is somthing that should also be tested.


Also thx for doing these tests.


edit: nvm, my point was already discussed and i missed it... shame that theres a lag. Looks like ps1 version is best.

[Tiberious]

Could some of the emulator delay be because of the CPS2 being misemulated at 12Hz as opposed to 8?

Think about this... if the game has its next frame ready to go before the input is, when it registers the button as being pressed won't exactly correlate with when you THINK it's been hit.

[Axel Kelly]

edit: nvm, my point was already discussed and i missed it... shame that theres a lag. Looks like ps1 version is best.

With all the weird stuff in the PS1 version (CPS1 chains??) in addition to the fact that no one will notice a 1 frame lag, I'd say DC still reigns supreme

[pdk]

Most surprising thing to me here is that the arcade version doesn't respond for four frames. What's the deal with that? Shouldn't it respond on the next frame?

consider the startup time inherent in many attacks; frames don't necessarily change on the frame that the button is pressed but it's the easiest thing to check for reliably unless you could take a look at the system's ram during each frame somehow

[Taskmaster]

Yeah, my timing's always off with CCC2. I only play iut when a friend brings it around.

NKI, did you try this for any other moves ( specials/supers) and if so, were the results the same?

[R-Jive]

^I'm that friend who brings it around, and yea we practically only play it at those occasions, It's really noticeable, thanks for your work NKI.

[Silks]

Sweet. Thanks NKI

EDIT: Just quickly glanced at the source files. CCC2 is strange as hell. If you look at Dhalsim's throw, the initial animation starts off 4 frames after arcade, but then the throw on the CCC2 seems to slowly catch up the frame lag. Watch each frame of ANIMATION of the throw change after both started, and CCC2 gets one frame less behind each change. So the throw animation on CCC2 plays through from beginning to end faster, but starts up on the 4th frame later. Almost like the game is trying to hide the latency. Wierd.

edit 2: actually, it only seemed to happen on the ccc2 left, arcade right vid. So right there, an interesting speed inconsistency in CCC2 :\

[Hanzo_Hasashi]

Props to NKI for this test!

[EveryFlowerFlow]

wait what? I thought CCC2's "lag" was baseless and people were whining for no reason.






=)



GS Mr.Nki

[JubeiNinja69]

funny i always thought the sound was the thing that was lagging in ccc2.

[Magnetro]

were your tvs the same??????????

[NKI]

With all the weird stuff in the PS1 version (CPS1 chains??) in addition to the fact that no one will notice a 1 frame lag, I'd say DC still reigns supremeI'd agree with this.

NKI, did you try this for any other moves ( specials/supers) and if so, were the results the same?I only tested normals and throws because I literally only had one finger free while recording, and because I didn't see a need to test every type of move. They all get their inputs just the same.

EDIT: Just quickly glanced at the source files. CCC2 is strange as hell. If you look at Dhalsim's throw, the initial animation starts off 4 frames after arcade, but then the throw on the CCC2 seems to slowly catch up the frame lag.That's because CCC2 runs at whatever crazy random speed it wants to. (It is well documented that CCC2's speeds are wrong. (http://forums.shoryuken.com/showthread.php?t=136571)) Just make sure you keep in mind that the speed of the game does not affect how long it takes a move to START...it only affects how long it takes the move to FINISH.

To illustrate this more clearly...

Let's say you had two arcade boards running side-by-side like I did in my test. Set board #1 to Turbo 0 and board #2 to Turbo 3, and use a Y-splitter to control both at the same time. Both boards will START their animation on the same frame, but board #2 will FINISH first.

were your tvs the same??????????(I'll treat this like a real question because surely there are some people who don't know you're being sarcastic.)

No, my TV's were not the exact same. That's why I repeated the test two times: once with the arcade version on the left TV, and once with the arcade version on the right TV.

[Ouroborus]

no wonder it felt weird playing mvc2 on a dc with a stick and converter compared to the arcade and dc pad.

[TaiPing]

Nice video NKI. Which is the best console port of ST so far in your opinion or the Evolution should use?

[Silks]

nki: I know what you're saying. I just wanted to point out my observation. Just more proof that the game speed constantly shifts from moment to moment; ie. it's not just faster or slower than arcade consistently a whole round. What game speeds did you have CCC2 and the arcade on, just out of curiosity? Cuz besides noticing that CCC2 played the throw animation faster than arcade on one vid, it played the animation one frame slower on the other vid. Whatever clock is running the CCC2 speed is wack as hell, constantly speeding up and slowing down. Which means, even input latency aside, the game is disappointingly too flawed for any serious player :\

(disclaimer: with how much heat I noticed between certain parties on this issue, let me state that I'm not some asshole blaming Sirlin. Like I said in the other thread, he was let down just as much as us, being that he's a player also)

[Master Chibi]

the music for this is totally best part of the vid

(sorry, it is, can't help it <3)

[deadfrog]

the music for this is totally best part of the vid

(sorry, it is, can't help it <3)

Aw dude, I was bobbing my head the whole time. :rofl:

Thanks a ton for putting this stuff to fair trial, NKI.

[Xenozip.]

Ah, I see. Well, I'm not sure what's going on with the emulators, but obviously they do not behave the same way as the actual arcade board. If I cared enough, I could test out the input lag on emulators running in real-time (which I think is much more relevant than emulators stepping frame-by-frame), but unless Evo decides to start using Kawaks for Evo, I really can't be bothered to wire up all that stuff again.
I can save you the trouble, I have completed various tests of my own, using several different emulators and games. Both real-time tests and frame advancement tests.

This has been posted in other threads, but I guess I should repeat it here since the topic resurfaced:

Final Burn Alpha, Mame, Mame++, and Kawaks have additional input delay for CPS2 games with audio enabled. The added input delay is a 2-frame lag.

Final Burn (the original version), and Nebula do not have this additional input delay.

This ONLY applies to CPS2, it does not apply to SNK games. And this only applies with audio on. If audio is disabled then there will be no added emulation-based input lag. The amount of added input delay from FBA/Mame/Kawaks for CPS2 audio is 2 frames.

With Final Burn and Nebula frame advancement tests, it shows that if you input a command the action will begin on the next frame, so it occurs on the 2nd frame. With FBA/Mame/Kawaks frame advancement tests it show that you can input an action and it will not occur for another three frames, so it occurs on the 4th frame.

With my real-time tests I was able to put PC games like MeltyBlood, GuiltyGear, and two of the emulators on the screen at the same time, and control multiple games at the same time. With this setup, I was able to show that actions were normally occurring on the same frame, but FBA/Mame/Kawaks lagged behind by generally 2 frames (though sometimes only 1).

Final note: If you enter the game's input test mode located in the dip switches, you can clearly see exactly what frame the input is being read by the emulator. With Final Burn, when you push a button, the input-test mode displays that input as being read on the same frame (1st frame). But actions do not occur until the next frame in-game (2nd frame). So this test shows that it's not the controller or the emulator that is lagging, it shows that the game recognizes the input instantly (same frame) but the action does not occur until the next frame. With FBA/Mame/Kawaks however, there is still the added input delay.

Could some of the emulator delay be because of the CPS2 being misemulated at 12Hz as opposed to 8? It appears to be directly related to sound, and only CPS2, so I imagine it's caused by poor Q-Sound emulation.

[Zaelar]

Even watching the source at normal speed you can see there is a delay. 4 frames...I didn't think it was this bad.

[NKI]

Nice video NKI. Which is the best console port of ST so far in your opinion or the Evolution should use?I still say the DC version is the best, but hopefully STHD will be out in time, and we can all just start playing that instead.

[SaBrE]

awesome test, and thats pretty wild. but what irks me is arcade st doesnt start animating til frame 4? that cant be right. are you sure the video converter on your supergun doesnt lag? most of em do =(

dont get me wrong, not trying to discredit the test, cuz it still clearly shows ccc2 has crazy lag. but if your video upscaler in the super gun lags, then that means ccc2 lags more than 4 frames, compared to arcade =/.

also, was ccc2 tested on a ps2 or the ps3? i wonder how much xbox ccc2 lags? probably just as bad im guessing since its same emulation. not like it matters anymore

[polarity]

If there's someone here with a supergun that outputs straight RGB (probably someone with a PAL TV with SCART, or an old RGB monitor) then we could eliminate that possibility.

[KMP]

awesome test, and thats pretty wild. but what irks me is arcade st doesnt start animating til frame 4? that cant be right.

That's simply the move's animation, not lag. The throws occurs after 1frame, the move, 3frames. There is simply 2 frames of his crouching roundhouse which have the same animation as the idle stance.

[World Warrior]

Slightly off-topic, does anyone know if there are any input lag issues on the CPS-1 SF games on CCC1?

[UltraDavid]

Is that really true? Do any moves start their animation immediately then? It seems odd that both moves NKI chose just happen to start their animation 4 frames after the input, seems more likely to me that the game just doesn't respond until that fourth frame.

[polarity]

That's simply the move's animation, not lag. The throws occurs after 1frame, the move, 3frames. There is simply 2 frames of his crouching roundhouse which have the same animation as the idle stance.

Uh, no, the same delay applies to throws too.

[KMP]

Is that really true? Do any moves start their animation immediately then? It seems odd that both moves NKI chose just happen to start their animation 4 frames after the input, seems more likely to me that the game just doesn't respond until that fourth frame.

There is nothing in the video that says Sim's throw starts at frame 4, you just assumed it did because NKI didn't count the frames for the throw. NKI should have used a LED for test #2 too.

Most command and special moves starts sooner than normal moves. I have to assume it's just delay to compensate the speed of the move since if you remove the so called delay Ryu's jab is 3 frames (1frame input delay + 2 frame move).

Edit:

Uh, no, the same delay applies to throws too.

If this is true, then the emulation is weird. That means command,special and throws are faster than arcade yet normal moves aren't. I have my doubts, but if NKI could do the LED test with Dhalsim's throw and counts the delay then everything would be cleared.

[Ouroborus]

hey nki, what was the song used for the magneto video?

[Silks]

How long it takes the move to start on arcade doesn't really matter in this case. It's considered our zero reference point. Also, like someone said, perhaps the moves don't start animating right away, but are considered to be active?

[Mechanica]

I still say the DC version is the best, but hopefully STHD will be out in time, and we can all just start playing that instead.

How would that work? Isn't that a 360-downloadable thing? Are there even useable 360 sticks?

[ramza]

How would that work? Isn't that a 360-downloadable thing? Are there even useable 360 sticks?

easiest would probably be ps3 w/ pelican ps3>ps2 converters

[Ubersaurus]

How would that work? Isn't that a 360-downloadable thing? Are there even useable 360 sticks?

360 has the Hori sticks. And being USB based I can't imagine it being too hard to get more arcade style ones hooked to it.

[Preppy]

MASSystems now can build you xbox 360 sticks too. I think most non-big-time-commercial stick vendors have solved the problem by just repurposing the control board from an original xbox 360 controller. End of tangent.

Great work, Sherlock NKI.

[EvilSamurai]

MASSystems now can build you xbox 360 sticks too. I think most non-big-time-commercial stick vendors have solved the problem by just repurposing the control board from an original xbox 360 controller. End of tangent.

Great work, Sherlock NKI.

We need a HRAP 360.

[ramza]

STHD IS coming out on PS3 as well...right??

[Onslaught2000]

Ramza: Yes it is

[Taskmaster]

I've seen 360 sticks in the stick thread. PS3 would be good, since everyone could use their current sticks. Good luck gettting people to provide PS3s for tourneys though.

[NKI]

awesome test, and thats pretty wild. but what irks me is arcade st doesnt start animating til frame 4?I tested out three other games on three other systems tonight, and here's what I got...

Street Fighter Alpha 2 Gold (PS1) - 3 to 4 frames
MvC2 (Dreamcast) - 4 to 5 frames
Hyper SFII Anniversary Edition (PS2) - 5 to 6 frames

And then if you look at the games I originally tested:

Super Turbo (arcade) - 4 to 5 frames
Super Turbo (PS1) - 4 to 5 frames
Super Turbo (DC) - 5 to 6 frames
Super Turbo (CCC2) - 7 to 8 frames

I dunno...4 to 5 frames for arcade ST sounds just about right to me...

Slightly off-topic, does anyone know if there are any input lag issues on the CPS-1 SF games on CCC1?If I had a copy, I would have tested it tonight...:sad:

The throws occur after 1frame, the move, 3frames. There is simply 2 frames of his crouching roundhouse which have the same animation as the idle stance.This is incorrect. Both moves occur (they begin animation) on the 4th frame. I double-checked Sim's throw tonight using the LED. Also, the first frame of Ryu's cr.RH is distinctly different from his ducking animation. (That's why I used cr.RH.)

NKI should have used a LED for test #2 too.The LED was irrelevant for test #2. I was using a Y-splitter, so both games got their input at the same time, and all we care about is if both games react at the same time.

Most command and special moves starts sooner than normal moves.Perhaps you're confusing hit frames with actual animation. The animation for Ryu's cr.RH starts 4 frames after the button is pushed, but it doesn't actually hit until 9 frames after the button is pushed.

hey nki, what was the song used for the magneto video?Sorry, I forgot to update the song list on my page when I put the vid up. (I've since updated the song list.) For the Magneto vid, I used "One Million Miles Away", by J. Ralph.

[Silks]

NKI: oh, so Sim's throw takes 4 frames until the animation starts after the input too?

[Hanzo_Hasashi]

If its not much to ask, could you test SF3 3S on DC too? That way you could kill two birds in one shot, clearing the 3s DC lag issue.

Sure, not having an 3S board to compare would make it hard but we can at least have how many frames it does lag after a button is pressed.

thx again for your time.

[Keits]

Or just compare DC 3s vs PS2 3s to see if there is a difference.

[SweetJohnnyV]

I dunno...4 to 5 frames for arcade ST sounds just about right to me...


This actually does seem a bit weird to me. Not that it has anything to do with the CCC2 lag, since the relative lag is obviously what counts. But I am kind of curious why the moves on the arcade wouldn't start up for ~4 frames.

I wonder if you were to do a SRK, or some other move that hits on the first frame, if it's animation would also lag by 4 frames. But I'm guessing that'd be pretty hard to test with your setup tho, wouldn't it? Oh well, just academic curiosity at this point...

[laugh]

First of all, thanks for the video and all the work you put into it, NKI.

Now, I have to point out something that can lead to some very bad misunderstanding. The way NKI presents the number of frames it takes for a move to come out is very misleading. Here's why.

In his tests, the arcade version showed the first frame of animation 3 frames after the button is pressed, NOT 4. He says 4 because he counts the very frame when the button was pressed. You can't count that into the number of frames it takes to animate a move, that's the 0-point not frame 1.

Also as a side note, we now know from lagless emulators that arcade ST shows the first frame of animation on the 2nd frame after the button was pressed. In NKI's tests, arcade ST running on supernova showed the first frame of animation on the 3rd frame after the button press, so there's 1 frame of additional delay on arcade ST board running on supernova when compared to the emulator. I can record footages from an arcade cabinet running an ST board (using the same setup: LED, 60fps) and see if the additional 1 frame delay was caused by supernova or not.

So in NKI's 4 frames this is what each frame means.

1 - the frame when the button is pressed, which means the game now recognizes that there's a change in input (This should really be frame 0)
2 - nothing shown yet
3 - nothing shown yet
4 - the game finished computing the change and is now showing the first animation frame of the move

I'm guessing maybe the real hardware has 1 frame of lag or supernova added that additional frame to before the first animation, because lagless emulators would show you that arcade ST infact works like the following.

0 - button is pressed. The game recognizes it.
1 - this frame doesn't show the change in animation yet, but the game is computing the change in input.
2 - first frame of animation shown.

[NKI]

NKI: oh, so Sim's throw takes 4 frames until the animation starts after the input too?Correct - same amount of time it takes all moves to start animating.

If its not much to ask, could you test SF3 3S on DC too?Hrm...I guess so. But I don't know 3S very well, so you'd have to tell me what moves are instant. All I know is that throws are not instant.

You can't count that into the number of frames it takes to animate a move, that's the 0-point not frame 1.Fair enough. Sorry for the confusion.

Also as a side note, we now know from lagless emulators that arcade ST shows the first frame of animation on the 2nd frame after the button was pressed.I'm not disputing the results of the tests using frame-stepping, but I'd want to test this with an emulator running in real time also, 'cause that's what really matters.

I can record footages from an arcade cabinet running an ST board (using the same setup: LED, 60fps) and see if the additional 1 frame delay was caused by supernova or not.Please do. I'll see if I can rig up a way to test emulators running full-speed.

[Xenozip.]

Also as a side note, we now know from lagless emulators that arcade ST shows the first frame of animation on the 2nd frame after the button was pressed.
Well it should also be noted that with the lagless emulator, both the emulator and game recognize that you have input a button on 0-frame.

In-game this is what happens:

0-Frame: You pushed a button
1-Frame: Nothing
2-Frame: Action begins

In the dipswitch menu under input test mode this is what happens:

0-Frame: You pushed a button and the indicator lights up.

When testing real-time with an LED, that could help show exactly how much delay there is between the LED lighting up and the game responding to a button being pushed. Since with a lagless emulator the indicator lights up the instant the input is recognized.

[KMP]

The LED was irrelevant for test #2. I was using a Y-splitter, so both games got their input at the same time, and all we care about is if both games react at the same time.

Yes, I know. I said that not because of the test results, but to know the real delay of the throw because my results were different.


Perhaps you're confusing hit frames with actual animation. The animation for Ryu's cr.RH starts 4 frames after the button is pushed, but it doesn't actually hit until 9 frames after the button is pushed.I'm not confused, I had to assume that from my results (which came out wrong) since it seemed really weird.
I retested it with another emulator (Kawaks) and the results are like the arcade. The wrong results are probably due to the fact I don't know if Nebula uses frameskip and how to disable it. I should have tested thoroughly.

[Xenozip.]

Kawaks has added input delay.

[KMP]

Kawaks has added input delay.

Yup I read your report on it, it made me realize emulators weren't perfect, thanks. But I wasn't testing for arcade perfect delay, only for nonrandom result (Which I got with Nebula).

[Digitalbooty]

Of course none of this is tested, yet, but it may be possible that superguns may have one frame of delay? That's crazy. I wish I had an ST board. I'd test it, too. I've been running ST on Nebula because it has what seems to be no input delay, but I guess there's a small possibility that the actual arcade board has one frame of delay, lol.

[Xenozip.]

Seems to me like the original Final Burn (not Alpha) is the most consistent with no input delay.

[Master Bigode]

Just made some tests, and found out that Ryu's c.roundhouse takes 5 frames to come out in MAME.
...Of course, this could be wrong since there's sprite drawing delay in CPS2 games running in MAME.
Hmmmmm....

[laugh]

In the dipswitch menu under input test mode this is what happens:

0-Frame: You pushed a button and the indicator lights up.


Actually, that's not what happens with input test mode. This is.

0 - Button pressed, game rocognized the input.
1 - Change shown on screen

There's no way you can show the change on the screen on the same frame the button was pressed. The fastest anything can happen is on the very next frame, which is the very first frame after the input.

[Xenozip.]

.... Ok, but that isn't what happens in the dipswitch mode with ST or SFA3. Though, that is what happens in-game and also in the dipswitches with Samurai Shodown 4.

Here's what I do.

Load Final Burn (original). Load Super Street Fighter 2 Turbo. [F2] to enter dip switches. (LP) to enter Input test mode. [Alt]+[P] to pause the game.

Hold a button, [Alt]+[O] to advance a frame, indicator switches from 0 to 1. There was no input being held on the previous frame, the first frame that the input occurs the indicator lights up.

Load up Final Burn Alpha. Disable Audio. Load Street Fighter Alpha 3 (980904). [F2] to enter dip switches. (LP) to enter Input test mode. [Pause] key to pause the game.

Hold a button, [F1] to advance a frame, indicator switches from 0 to 1. There was no input being held on the previous frame, the first frame that the input occurs the indicator lights up.

In Samurai Shodown 4 it is as you describe.

Load up Final Burn Alpha. Disable Audio. Load SS:IV. [F2] to enter dip switches. (A) to enter Hardware Test mode. (Start)x2 to advance to the input test mode. [Pause] key to pause the game.

Hold a button, [F1] to advance a frame, indicator stays unlit, advance another frame and the indicator switches from 0 to 1. So the indicator switches on the second frame of the input being held.

-- In-game is another story. For SFA3 you can pause, hold an input, advance a frame and nothing happens, advance another frame and the action occurs (so it occurs on frame #2, the second frame the input was held).

Now, if you want me to verify this with a live real-time capture then I probably can. But this subject is really rather irrelevant.

[laugh]

.... Ok, but that isn't what happens in the dipswitch mode with ST or SFA3. Though, that is what happens in-game and also in the dipswitches with Samurai Shodown 4.

Here's what I do.

Load Final Burn (original). Load Super Street Fighter 2 Turbo. [F2] to enter dip switches. (LP) to enter Input test mode. [Alt]+[P] to pause the game.

Hold a button, [Alt]+[O] to advance a frame, indicator switches from 0 to 1. There was no input being held on the previous frame, the first frame that the input occurs the indicator lights up.

Load up Final Burn Alpha. Disable Audio. Load Street Fighter Alpha 3 (980904). [F2] to enter dip switches. (LP) to enter Input test mode. [Pause] key to pause the game.

Hold a button, [F1] to advance a frame, indicator switches from 0 to 1. There was no input being held on the previous frame, the first frame that the input occurs the indicator lights up.

In Samurai Shodown 4 it is as you describe.

Load up Final Burn Alpha. Disable Audio. Load SS:IV. [F2] to enter dip switches. (A) to enter Hardware Test mode. (Start)x2 to advance to the input test mode. [Pause] key to pause the game.

Hold a button, [F1] to advance a frame, indicator stays unlit, advance another frame and the indicator switches from 0 to 1. So the indicator switches on the second frame of the input being held.

-- In-game is another story. For SFA3 you can pause, hold an input, advance a frame and nothing happens, advance another frame and the action occurs (so it occurs on frame #2, the second frame the input was held).

Now, if you want me to verify this with a live real-time capture then I probably can. But this subject is really rather irrelevant.

I think you and I are talking about the same thing with slightly different understandings. I know exactly what happens in input test mode on FBA without sound and FB with sound etc. I've done the exact test you just described many times before, and I'm sure you did too.

I think you believe the game recognizes the input on the frame AFTER you press frame step while holding the button, and I believe the game recognizes the input on the frame BEFORE you press the frame step for the first time. I think like that is this is my educated guess of what happens in a frame of time in sequential order.

1. [update the screen and sound prepared during the previous frame] (which I believe is the moment you are stuck at if you press pause or frame step)

2. [check the controls for change in input for both players]

3. [computations]

4. [apply the changes according to user inputs on the next frame if possible]

I say "if possible" because some things take more than a frame to finish processing (ie. mid-fight computations) and some things can be processed within a frame of time (ie. changing the character portrait's color scheme according to the button the player chose to pick the character on the character selection screen, and changing a 0 to a 1 or vice versa in input test mode). It all depends on how many operations capcom decided to run on certain changes in inputs.

Simply, I don't think it can recognize the change in input and display the changes onto the screen at the same time. Obviously, the game must have known it had to make changes on the screen before the moment came to send the updated video signals.

Sorry to derail from the topic.

[Xenozip.]

Well I do understand that it doesn't make sense from a programming perspective. But on the other hand I'd still like to do some further testing to verify.

[Retekin]

Curious about the DC vs other versions of Street Fighter 3 thing, if memory serves Chun's crouch jab has 1 frame startup? Try using that. If you don't really want to be overly scientific about it though, just so long as you test the same move and just count the difference it should be fine.

Edit: also, thanks for going into such detail on this.

[Ali Vegaz]

I think everyone just just play this shit on arcade... this isn't something that you should toy with on consoles n shit... There are speed issues... shit, now that I look back, it probably wasn't even the speed that was fucking me up, it was the input lag... wow... this should be an arcade only tourny, if no supergun/cabinet is available then it should be ran off Xbox Mame or PC Mame... period.

[SaBrE]

eh. emulators lag too, ESPECIALLY mame. and the xbox emulators are trash for the same reasons. evo will not put st on supergun even if they had 50 superguns and boards with absolutely perfect sticks(or able to plug your own in) and video converters that dont lag. reason being, is if we put st on arcade, then everyone else will want their game on arcade. and we arent doing the arcade thing anymore. cant do emulators anyway since of all the legal matters, and the sponsors dealing with us. console is the only way, period.

[Toodles]

I think everyone just just play this shit on arcade... this isn't something that you should toy with on consoles n shit... There are speed issues... shit, now that I look back, it probably wasn't even the speed that was fucking me up, it was the input lag... wow... this should be an arcade only tourny, if no supergun/cabinet is available then it should be ran off Xbox Mame or PC Mame... period.
And when I dream, I have a pony.

[dublo7]

well that was great work again NKI :rock:. i have just noticed this timing craziness in CCC2 @ NEC. however i noticed it btw 2 _different_ (?) unmodded ps2 machines. that seemed really wierd. But i do play my us CCC2 on a modded ps2. can a modded ps2 influence lag time on CCC2?

boy do i sound greedy.:confused:...but its just a question for the next research grant....:wink:

[King Of Bums]

I remember looking at the tekken 5 arcade machine, and i notice they have a thing at the side where you can put in your own controller. Somebody needs to do that for all the games at evo. Then the arcade junkies, pad players, and custom stick guys will be at peace with each other at last. Also, program in an option so you can set the buttons the way you want. Its never going to happen, but its a great idea.

[polarity]

actually it's a terrible idea, because feasibility is kind of a prerequisite for an idea to be considered great. anyone can come up with some ridiculous, ideal solution that's never going to happen.

[Xenozip.]

What Polarity means is, the only reason console is used in the first place is because there are no arcade cabinets available.

So what if the cabinets had a controller socket? We still wouldn't have any cabinets. Unless you want to buy Evo all the cabinets they would need to run a tournament..

[Mechanica]

Stopped reading at the top of this page but don't bother with 3s. There's no reason to test it, you don't need to test it to know how fucking awful of a version it is. It's not so much about input lag as it is the juggle/combo/unblockable system. For a much lower level player it's not an issue, if that's what the poster was worried about.

[Henaki]

Stopped reading at the top of this page but don't bother with 3s. There's no reason to test it, you don't need to test it to know how fucking awful of a version it is. It's not so much about input lag as it is the juggle/combo/unblockable system. For a much lower level player it's not an issue, if that's what the poster was worried about.

Actually the slight speed up just causes it to ignore the occasional frame which is what fucks stuff up.

Unless you mean the DC version? In which case yeah being based on version B is what fucked it up.

[elvis_a_presley]

eh. emulators lag too, ESPECIALLY mame.
On Windows/Mac/Linux+X-Windows, yes.

On Linux with a low-level library that can access framebuffer hardware directly and create arcade-perfect modelines, there is no lag (or at least it will be within the "0-1 frames" range). AdvanceMAME (http://advancemame.sourceforge.net/readme.html) + Linux + SVGALib allows writing directly to video card farmebuffers. Coupled with Linux's USB support (which has been measured to be lag free compared to Windows and Mac) it becomes closer to real hardware than any console.

Everyone who complains about emulation lag plays emulators through high-level code like a Windows desktop. When there's a dozen abstraction layers of code between you and your hardware, of course there's going to be lag. When you understand how SVGALib works, you understand why AdvanceMAME+SVGALib provides a lag-free option. The downside is it's not a trivial thing to set up. Most people I've seen either try and fail, or set it up wrong. It takes a little understanding of how an arcade monitor works (http://easymamecab.mameworld.net/html/monitor1.htm) to set it up correctly.

I'm keen to prove that statement too. I only have access to a 50Hz camera (PAL territory) at the moment, but I know a few guys with access to proper 60 fields per second cameras. If I can borrow their gear, I'd like to test my Linux/AdvanceMAME/SVGALib setup using NKI's methodology and prove my above statement scientifically.

As for the legality of emulators - there are plenty of people selling legal ROMs:
http://www.dreamarcades.com/cabiocade.shtml

Capcom are not anti-emulation, nor anti-ROM. You'd need to present a decent business case to them, but if it was done right I'm certain you could get them to license legal ROMs to you.

[jugoso]

I use sdlmame, but the problem is not input lag/delay but audio delay

[elvis_a_presley]

I use sdlmame, but the problem is not input lag/delay but audio delay

SDLMAME's output goes through more steps to get to your screen/speakers than AdvanceMAME. Ditto for standard Windows MAME.

As I mentioned above, AdvanceMAME is the only MAME build that is allowed direct access to hardware. The downside being it's difficult to set up properly (you can still run it under Windows or X, but that's no better than normal MAME), and the fact that the author has stopped doing work on it, so no recent builds are available.

[ShinAkuma204]

Well, I am very hesitant to say this, because I really want the DC version's input lag to get fixed in STHD, but...it's not a huge deal. Most people (myself included) probably can't detect one frame of input lag just by playing the game, especially when the lag isn't constant. (Remember, sometimes the DC version has no lag.)

Just one question NKI - did you use a DC hacked pad for test? The DC pads triggers (analog I believe) are supposed to have some input delay over the other four digital buttons. Since one of the triggers would normally be mapped to the roundhouse that could explain why the DC sometimes lags.

It's possible the DC has no lag and it is simply the controller.

EDIT: My point seems to have been mentioned earlier in the thread.

[NKI]

The DC pads triggers (analog I believe) are supposed to have some input delay over the other four digital buttons.

I made sure to use the face buttons (not the triggers) to avoid this exact problem. It still lagged one frame every once in a while.


what video outputs did u use?I used the SuperNOVA's AV output.

further more are you counting 29/30 frames per sec

or 59/60 frames/fields a sec?59/60.

[PozerWolf]

Sorry, I forgot to update the song list on my page when I put the vid up. (I've since updated the song list.) For the Magneto vid, I used "One Million Miles Away", by J. Ralph.
Ah, thanks for the info. Song was too good.
Great video, too.