GGXXAC+ Input Lag Testing - Results

[Teyah]

Hi guys - you might recall that I mentioned that I would be testing the new PS3/360 ports of AC+ and comparing them to arcade/PS2 AC. Well, I finally got around to dusting off my PS3 and testing that version and finishing off the tests, which you can see for yourself below, in .gif form. I might compile these into a video in the future, but for now I just thought I'd let you guys know where the different versions stand in terms of input delay.

First of all I'd like to mention the methods I used. I used the red LED light on a TE Kitty to time the input of a button press to the exact frame, then counted the number of frames until the first frame of animation, using a 60 fps camera to capture video. I used Faust's 5D as the button input, since it has a huge first-frame animation (the huge tornado spin). I took several passes of video for all versions of each port and confirmed that the results shown below were, by far, the most common result. Roughly 90% of the time, the results were as shown below, the other 10% of the time they were one frame slower - this is likely due to variances in input timings (whether the input arrives at the beginning or end of the frame buffer). After taking roughly 100 tests per port, I can say with confidence that these input delay results are accurate.

Here's a list of the equipment that I used:

Canon PowerShot SD800IS - used for taking 60 fps video

Panasonic CRT TV - Panasonic is known (and tested by me) to have the lowest lag CRTs on the market, especially when it comes to HD CRTs. This CRT was confirmed to be 100% lagless. I've also tested 360 and PS3 AC+ on my Panasonic HD CRT TV with the same results as shown below.

Samsung SyncMaster 955 DF - a zero-delay CRT computer monitor, used to test AC arcade and #R AC, as well as 360 AC+

PC running Demul @ 60 fps - used to test AC Arcade input delay

PS2, PS3, Xbox 360 consoles - for testing GGXXAC+

Now for the results. Watch the red LED light to see when the zero frame of input is.

Fastest: Delay 2

GGXXAC Naomi (Arcade) version - emulated via Demul

The arcade version runs with a native delay of 2 frames via emulation. This is consistent with other players' findings - Elvenshadow and others have mentioned that they can immediately feel the difference when switching from arcade AC to console AC, and this is most likely the cause. The arcade version runs the fastest out of all of them.

GGXX #Reload PC version

The PC release of #Reload has been very popular in online circles and now it's not hard to see why - it has only 2 frames of native input delay. As the only existing port of the GG series that has been able to match the input delay of the arcade, playing online on #R PC can feel seamless. At an in-game online delay setting of 2, which allows you to play players within close proximity, it matches the delay of the other console versions (except PS3). Offline, it's a 100% accurate arcade port, as far as input delay is concerned - which is something that no other port of GG can say.

Next fastest: Delay 3

GGXXAC+ PS3

The PS3 version of AC+R runs at a native delay setting of 3, which falls in between arcade AC and console (PS2) AC. This is a new standard and might be tough for players of other versions to adapt to - it certainly was for me, as I was dropping a ton of Millia air combos on my first set of matches on AC+ PS3 on a CRT. However, if you're playing on a low-latency LCD screen, the total delay will end up being 4 to 5 frames, because of additional LCD delay. In this case, it's much better to choose the PS3 version, as you would then be playing in a delay setting similar to PS2.

For online play, while the PS3 version beats out the 360 in terms of offline delay, the PS3's less than stellar online play reduces this advantage that this version has over the 360. In the end it's tough to say which version is better for online. If you know your opponent is not using wireless and plays close by, it would likely be best to play on PS3 though.

Slowest (the rest): Delay 4

GGXXAC+ PS2

Interestingly (and sadly), the PS2 version actually runs with 4 frames of delay as well. This may very well explain why things like 5D "only hit in America" - because "America" (console players) have 2 fewer frames to react, compared to arcade players! Thankfully, this will be getting fixed with AC+R, with 5Ds and other reactable things being slowed down to match the higher delay setting that AC+R uses.

Please note: the reason that the screen doesn't begin updating a frame earlier is due to the syncing of the camera and the TV. I did take multiple videos to test across multiple CRTs, but in this case, I forgot to turn off the camera between sets so they all sync up this way. Also, the LED does not turn on, because the LED functionality doesn't work on the PS2 connector. However if you eyeball the button press, you can see that it matches up with the other videos and is a reasonable basis for calculating delay - especially with over 100 trials being done to make sure the numbers are reliable.

GGXXAC+ Xbox 360

The 360 version of AC+R runs at the same delay setting as the PS2 version. I tested this one on the CRT computer monitor as well, with the same results of 4 frames of delay. Also, it may just be me and others I play with, but it feels as if the 360 version runs smoother than the PS3 version - it doesn't have any slowdown during gameplay. In this sense it would be "PS2-perfect" as the PS2 doesn't have slowdown either, though PS3 feels like it has some. I'm going to find a way to test this in the future, so take this info with a grain of salt for now.

Full gallery of results here: http://imgur.com/a/GFfs5#0

Now, the good news. GGXXAC+R will be coming in Spring 2013. This game has slowed down reactable overheads and other similar moves by 2 frames. As well, many players have noticed increased input delay which is the standard for current generation games (almost all run at delay 4 or above). So, it is a good assumption for now, that AC+R will increase from native delay 2 to native delay 4. This would make the online experience a lot smoother in terms of not being hit as easily by unreactable dusts and things of that nature, since the game itself is being changed to allow more time to block certain types of mixups. As console players, we shouldn't notice any difference at all, since we've been playing in delay 4 ever since the PS2 ages. We may even actually have a port that is less laggy than the arcade version, as arcade may be running at delay 4 while PS3 runs at delay 3. We'll just have to wait and see.

Please leave questions or comments below, I'll be happy to answer. Merry Christmas!

[WolfCrimson]

Wow, thanks for the info!

I have an observation though: in the gifs of Faust's 5D for the PS3, 360, emulated arcade and PC reload ports, you can see an "incomplete tornado" animation starting a frame (1/2 a frame??) earlier than the results you posted. Wouldn't that mean that the results should be corrected by 1 frame (1/2 a frame) in those ports, since that incomplete tornado is the earliest visual cue of the game reading your input (and hence is the 0 frame of 5D)? I don't understand why it isn't the case.

[Teyah]

The incomplete animation is the CRT rendering that frame - it doesn't finish until the next frame. This is typically how input delay results have been historically recorded - you count up to the completed frame in the delay count. Depending on the camera capture syncing (which I can't control, outside of taking multiple passes which I did), that incomplete frame may not have even shown up, and a good chunk of the time it doesn't. Sometimes, the "incomplete frame" is only captured for the top 20% of the screen, where there is no useful information at all (generally). Other times, the incomplete frame is skipped in recording and the total delay frames are the same as above. Either way, the refreshing of the frame isn't complete so it isn't counted. I actually made sure I included the updating frame in the above pics for comparability's sake - otherwise there would be the question of whether the frame shown is the updating frame - this way there's no confusion on that end.

[Amadeous]

Amazing, and thank you Teyah. Though Circuitous brought up an interesting point a while ago, apparently he feels that the "macro delay" isn't present in 360/PS3 ports? I dunno how you could test this; but if the alleged 3 frames of delay when using a macro to FRC aren't in the newer versions that'd be cool.

[WolfCrimson]

Ah OK, I see. Thanks.

[Sym_]

http://shoryuken.com/forum/index.php?threads/sub-1-frame-hdtv-monitor-input-lag-database.145141/

There are lcd monitors with virtually no lag.

[qwerty]

Good shit.

I guess now I know why I had such an easy time adjusting to the 360 port, scratching my head at people saying the PS3 port was hard to adjust to. Between the speed difference and reduced input lag, it would make sense that the PS3 version would be the hardest to adjust to.

[Silmerion]

Man. This explains why the PS3 port felt faster than the PS2 version did.

Science!

[Teyah]

http://shoryuken.com/forum/index.php?threads/sub-1-frame-hdtv-monitor-input-lag-database.145141/

There are lcd monitors with virtually no lag.

(edited)

All LCD monitors have lag

[Tong]

Nice findings, this sure clears some things up. Thanks for taking your time

And there must be something wrong with my #R PC, since it feels *really* laggy (8~9f delay at least). Same, but to a less extent, happens in AC+ emulated on Dolphin. The PS3 port is by far the closest to perfection.

I play Baiken, so if I had to describe how it feels to combo a j.P after Tatami it would be:

#R: Must press P, as I hold up, just as tatami recovers

AC+ (Wii): Press P after tatami is gone

AC+ (PS3): Press P during the actual jump, it comes out instantly.

Easiest part by far, problems happens when Airdashing after j.D is needed. The difference is such, that I had to re-work my whole muscle memory when playing on PS3

[BioLogIn]

Thanks, Teyah! Good to see these things sorted and checked in efficient manner )

The arcade version runs with a native delay of 2 frames via emulation. This is consistent with other players' findings - Elvenshadow and others have mentioned that they can immediately feel the difference when switching from arcade AC to console AC, and this is most likely the cause.

Sorry, I don't quite get what you've meant by quoting this, can you please elaborate? I mean, you check Demul running arcade game image, find that it has 2 frame input lag (likely introduced by emulator) - and then we off to Elvenshadow comparing Arcade version to PS2 version... but neither of this versions are the subject of this exact test (you give results of PS2 version test later).

Also which kind of connection to TV did you used in each (console) test? Composite? S-Video? Component? Composite to SCART even? )

Also, when we compare all this to an arcade version, do we assume that arcade version has 0 frames of input lag? Or do we allow a possibility of it having a 1 frame delay? Also, as far as I understand, it is kinda hard to differentiate between this two options, as the input can arrive in a different time of 16(6)ms window, both before, during and after actual frame processing - is there a certain way of differentiating this situations? Or do we simply assume that 16(6)ms window starts (or ends?) by processing the frame and work with this assumption?

[Teyah]

The lowest input lag possible in any game is 1 frame after input, since the result cannot occur on the zero frame (the game would have to be able to read your mind to do that). On top of this, there were previous findings (back in 2004-2005) showing that there is a 1 frame input buffer in GG. This would bring the lowest possible delay to 2F, which is in line with #R PC and emulated arcade results above. If there is any added delay by Demul it is negligible.

The way to differentiate between situations where the input arrives at the beginning of the input frame (0.0 ms) and at the end of the input frame (16.6 ms) is to simply perform a large number of trials and determine where the average falls. In the tests above, the large majority (85-95%) fell within the delay frames I've shown above.

I just mentioned the comparison made by Elvenshadow and other japanese players because it backs up my findings - that PS2 version runs with extra input delay compared to arcade, which people have reported previously.

All CRT tests were done using analog signals. Composite for all tests to CRT TVs, Component to HD CRT TVs. VGA was used for the CRT monitor. All CRTs were consistent with their results - eg. Xbox 360 GGAC+ showed Delay 4 when tested on all of regular CRT via composite, HD CRT via component, and CRT monitor via Xbox VGA cable, in roughly 90% of all trials.

[AtTheGates]

good to see that i'm not crazy after all, i immediately thought i felt that the ps3 version has more immediate input than the ps2 version, which was very surprising - i expected more delay (lcd monitor instead of crt, its a modern port, i use an adapter for my ps2 stick to ps3 etc). never thought i'd feel 1F that much, but it was really apparent with slayer.

i'm gonna stick to ps3 on tournaments now i guess.

edit:

So, it is a good assumption for now, that AC+R will increase from native delay 2 to native delay 4. This would make the online experience a lot smoother in terms of not being hit as easily by unreactable dusts and things of that nature, since the game itself is being changed to allow more time to block certain types of mixups.

if i understand you correctly, shouldn't the time be exactly the same? if we have 2f more of input lag, but 2f more time to block most overheads, shouldn't it equal 0? at least if we assume they generally add +2f to ps3 and x360, which would be 5f and 6f, which i somehow doubt. what they would have to do is add +1f to the ps3 version to get it up on par with the arcade and 360 version, and i also doubt they will do that.

[reaVer]

Teyah: what about nullDC with my plugin?

[Mitsurugi]

Thanks a lot Teyah (^_^)b

You even took the time to add some nice gifs to make it easily understandable by anyone !

Sorry, nothing really helpful here however one should not forget about thanking people who deserve it.

[Teyah]

if i understand you correctly, shouldn't the time be exactly the same? if we have 2f more of input lag, but 2f more time to block most overheads, shouldn't it equal 0? at least if we assume they generally add +2f to ps3 and x360, which would be 5f and 6f, which i somehow doubt. what they would have to do is add +1f to the ps3 version to get it up on par with the arcade and 360 version, and i also doubt they will do that.

Oh, the +2F refers to the arcade version only, which runs at 2F delay for AC but has likely been bumped up to 4F for ACR. I very much doubt that Arcsys would increase the delay of the console versions by 2F as well. I do have confidence that their team is aware of the input delay issue and would test the consoles to see how much delay they need to add to bring it in line with the arcade version of AC+R, to which the answer would be none (or maybe 1F to the PS3 version only).

reaVer: For some reason I wasn't able to get nullDC to run at a consistent 60 fps on my desktop computer, and the stick with the LED doesn't work with my laptop setup. I'll test it from my laptop with a different stick and try to eyeball the button press (like I did with the PS2) a little later.

Mitsurugi: Thanks!

[Blade]

Mind if I ask a theoretical question here?

People have frequently pointed out to me that they were experiencing lag in matches with me for varying reasons, yet I find it difficult to notice myself unless it's an extreme circumstance.

If say, someone like me who has been playing the PS2 version of Accent Core Plus for a very long time...and also the BlazBlue games online...

Is it possible to be "acclimated" to lag? Like, adjusted to it to the point that it doesn't effect matches?

[Silmerion]

Is it possible to be "acclimated" to lag? Like, adjusted to it to the point that it doesn't effect matches?

No matter how used to a given lag you are, you'll never be able to react to things as fast as you could in a relatively lagless environment.

[TheRealBobMan]

I can understand that. I can adjust to lag somewhat well at this point - I used to be a very mechanical player when I played Melee, learning to powershield projectiles to reflect them (same 2-frame window as Slashbacking, but with Melee's stupid 5-frame start up on blocking), and I had to adjust to varying amounts of TV lag all the time. It was frustrating as hell, but I'm somewhat used to it after all that practice. Now cue the lag with GG online - I can't ****ing stand it when I'm can't punish Riot Stomp on reaction because of the lag, but as long as the lag is consistent I can operate on prediction and still do decently. I mean, I've run into players online that ALWAYS run the same block string, and I can still inconsistently slashback the 2D I'm expecting at the end about 2-3 moves later, after a few games to get used to how big the delay is. I played against a Ky once where if I timed slashback on the first hit of Charge Stun Edge while waking up, I'd get the slashback on the 3rd hit. The only problem is when the delay feels really variable - if it keeps changing during a match, I can't get a handle on it, and I drop absolutely everything.

I don't know what to say about you not being able to notice it though. I'm sensitive to even really slight delays and adjustments in timing, even if I can adjust to them pretty well at this point. Silmerion is 100% correct though - you cannot react to some things because of the reduced time to react to them. Therefore, it will always affect what you could do in a lag free environment. You're probably just not trying to punish things that you should be able to punish in a lag free environment.

Anyway, thanks for the posts Teyah. Really good information there.

[qwerty]

Is it possible to be "acclimated" to lag? Like, adjusted to it to the point that it doesn't effect matches?

The thing is, it's not as simple as just "getting used to" things being different from platform to platform. It's one thing to slightly adjust your timing for tight links or FRC's or whatever, but that doesn't change the fact that the game is reading your inputs so many frames slower than it should be. It affects defense in particular as someone else already pointed out, because you can find yourself getting hit by things you reacted to. It also affects the neutral game, since it effectively makes pokes slower and therefore harder to punish with (not to mention that it can make hit confirming harder, but with this game that's less of an issue thanks to hit stop).

The only thing we can do as players is accept the fact that the game's gonna be played a bit differently depending on the platform. Just play like you normally would and adjust as needed.

[nstalkie]

Mind if I ask a theoretical question here?

People have frequently pointed out to me that they were experiencing lag in matches with me for varying reasons, yet I find it difficult to notice myself unless it's an extreme circumstance.

If say, someone like me who has been playing the PS2 version of Accent Core Plus for a very long time...and also the BlazBlue games online...

Is it possible to be "acclimated" to lag? Like, adjusted to it to the point that it doesn't effect matches?

Just to make sure we are talking about the same thing: this discussion is about input lag, not network lag. So if you play people online and they complain about lag: they are complaining about network lag. When you play people offline and they are complaining about lag: they are complaining about input lag. I think you may be confusing the two here. My apologies if that's not the case, but there's a huge difference between the two.

I've seen input lag used as an excuse so many times at tournaments. "I can't hit my links, setup is laggy as ****", other guy picks up the same stick and pulls of some one frame links from the first go: "no it's not" (seen this a couple of times). Many people are used to playing on laggy setups all the time at home, that when they actually play on lagless setups, they feel "lag". Another thing that frequently happens is that people get nervous at tournaments and start mashing without even realising it, making them miss things, then blaming it on "lag". Another frequent excuse is "stick malfunction" btw, but that's material for another topic This reminds me of the infamous moment when 3S was released on PS3. A guy on stream was trying the evo moment and kept failing for I think 45 minutes. He was complaining about the huge amount of lag on the PS3 setup. Then a Japanese pro player came on stream and tried it: got it on the second try (same setup, he was at the streamer's house). The Japanese player did say he felt there was slight input lag as opposed to arcade 3S, which he was used to play, however he had no trouble adapting it seems. Don't remember the name of the US player.

[BioLogIn]

When you play people offline and they are complaining about lag: they are complaining about input lag + video processing (mostly TV processing + possibly HDMI issues, depends) lag. And TV lag is likely to be can be much worse than input lag... if TV was chosen poorly.

[Teyah]

May as well add this in. I tested PS3 and 360 GGXXAC+ via HDMI on my HD CRT TV, both show the same Delay 3 / Delay 4 as found on the non-HD CRT tests via composite video. So HDMI doesn't inherently cause lag, from what I have seen.

[Chrome Homura]

Amazing, and thank you Teyah. Though Circuitous brought up an interesting point a while ago, apparently he feels that the "macro delay" isn't present in 360/PS3 ports? I dunno how you could test this; but if the alleged 3 frames of delay when using a macro to FRC aren't in the newer versions that'd be cool.

This is relevant to my interests. I got the same impression from the casuals I played at NEC, although to be fair I wasn't used to the PS3 speed... it might merely be that, but it'd be nice to know in any case.

[freelander]

and this year's Dustloop Christmas Science Fair Project winner is:

Teyah