No, seriously. DC is not 3S at all because there’s a global shift in startup/active/recovery frames. It’s the least accurate out of any iteration.
Hey there, congratulations for your persistence at running these tests, it is a good thing that someones tries to sort things out and get some facts !
I discovered shmupmame in your thread, and I was quite surprised…
Since you also mentioned Groovymame, I had a good look at the forum where the devs post.
Here is what I understood, roughly :
-
Shmupmame is very easy to use, and reduces input lag by breaking the process of emulation, which can result in, mainly, visual inaccuracies (but so far I couldn’t spot anything except on random screenshots). The other downside being that you can’t use Vsync or tripple buffering if you need it to get rid of screen tearing… well, not completely sure of that part.
It also offers a nice way of mimicking the arcade looks using HLSL.
I fiddled with it for a few weeks now, trying to check it out a bit after spending some time playing at the arcade. But I don’t know if it does add lag or not. (it may not be the case as the various hacks of this emulator just cut the cue of rendering to get straight to business, so this may be something to test).
I’ll try to edit with screenshots later on, even though my HLSL settings are not there yet. -
GroovyMame changes the way emulators work, and the specific order in which they do run the steps of emulating, making it accurate, but built in a more logical way, perfomance wise.
Bottom line, so far, both emulators seem to get the same results regarding frames of delay.
But, GroovyMame devs seem very responsive and striving to get to the point where you won’t have any lag coming from the emulator (while flagging the outside causes and trying to find solutions) while remaining accurate. Great project. But seems like a pain in the ass to use for now. so we can hope that at some point they’ll get a guy and more things to make it user friendly.
So right now I’m using Shmupmame, and even though I do have a laggy TV, it’s getting damn close to the arcade, IMAHO.
Hope some of this makes sense.
I’ll add links to threads worth reading if you want to check the detail of what I tried to explain there, later on.
[EDIT] :::
-
Here a link to Papasi’s thread here on SRK where he made all the same lag tests and more, specifically on ST using his PCB and Shmupmame as comparison.
-
Somwhere along the thread, Jdubs came in and linked to a thread where he started to talk to GroovyMame’s devs about input lag. They’ve been very responsive and tested a lot of thing and actually made some great discoveries that should largely improve their next release. Very long but interesting read. btw, I’d have to check again to confirm before spreading bad info, Seems like in this thread, Dark Gaiden has tested the influence of bilinear filtering and claims it adds some lag. But I can’t say if he tested it or if it’s a hunch, I’ll message him to see and edit later. Hence me posting screenshots below with and without this option turned on as it’s fairly common to use it to make the game lookable on a LCD (not needed at all with HLSL)
[EDIT 2] :::
- here is also a link to the official ShmupMame page, it is pretty short to read and understand the means the dev uses to get it going.
Can you also please post which input methods (and the brands/manufacturers) you tested? In my experience the MadCatz TE PCBs are slightly laggier than some other brands. I don’t know if they’re universally laggier across all platforms or just on PS3 though.
It’s pointless to argue over 1f of monitor/emulator input lag if your controller is just as laggy.
Sure, I used both an old crappy Hori Ex2, and a HRAP VX SA for Xbox360.
Using USB sticks on emulator seems to add 1 frame of lag, from what I read it should use keyboard material or some sort of bridge to get rid of it, but this is not my point. But it seems that the guy that works on GroovyMame is aware of the issue, and will ask official mamedevs to change it in future iterations to make implement RAW input for sticks too, not just keyboard.
I just meant to say Shmupmame (and most likely GroovyMame) are an improvement on FBA and old Mame.
Don’t get me wrong, I don’t mean to fuel a flame war of what’s good what’s not, like PC sucks or rocks.
I just think the more access we have to better versions of the game, the better.
And it seems like something cool aside console, which is good in other ways but also sucks.
As promised, here are screenshot comparisons of different graphic options on Schmupmame (or any emu right now I think) to display HLSL:
Sorry for the big image, spoiler tags don’t seem to work.
Please bear in mind I edited the HLSL settings more or less as I could, there’s slightly too much blue and red, I have to get back to the arcade to check again.
Still, it does look a lot more like arcade, and way better than the regular scnalines filter, which is great if you’re stuck with a LCD.
By the way, I mentionned Bilinear filtering as it seems that it also adds lag. Yet I have to check again to confirm.
I was actually referring to the tests performed by the OP, but thanks anyway.
Not sure what you mean by “input method,” but I listed almost everything I did in the original post. 360 was on hori pcb. Ps3 and emulator on mc Cthulhu.
Weird. For some reason my browser caches page 2 as page 1, so I only see the first post from page 2, not the original that listed the controllers. Anyway it’s sorted out now. I had to clear my browser cache and log in again.
I’m not sure what isotopez’ thoughts on this are - from my perspective, I think we could do it but it would be really low priority. it’s taken a good while to actually set up the speed tests (plus we both work so it hasn’t always been our top priority lol) and we plan on actually doing those pretty soon. we’ve thought about going back and re-testing for input lag with isotopez’ more accurate method after we’re done with the speed stuff, and we still might do that.
IMO Dreamcast and PS2 are kinda bottom of the importance list. there’s more to be gained from testing the big platforms that a lot of people use - CPS3, OE, and emulator. not saying we won’t do it, just won’t be at the top of the do list.
I doubt we’ll do anymore tests with the dreamcast. There was 1 thing I still need to find out for esn, but other than that, it seems bad enough that no one should be playing 3s on it. I don’t have access to a ps2 anymore, so we probably won’t run speed tests on it or retest the lag with a stock controller either.
So Lance and I finally got around to running our speed tests using our programmable controller.
I programmed the controller to execute:
- x6 elbow cannons on urien
- chun low forward -> super (canceled on the last possible frame)
- 123 -> genei jin link standing strong
- ken’s target combo (with the strong canceled on the last possible frame)
We ran all these macros with timings adjusted for the 60Hz that oe refreshes at and the 59.583Hz that cps3 refreshes at.
The purpose of the elbow cannons is to see how a long and timing specific combo is affected by the game’s speed. The other macros are mostly for testing the cancel windows between platforms i.e. maybe oe’s cancel windows are shorter and that’s why it feels different than arcade.
What we found was that the elbow cannons only work on 360 with 360 timing. They work on cps3 with cps3 timing, and they also sometimes work on cps3 with 360 timing.
The other three macros all worked with cps3 and 360 timing on both platforms.
Interesting find isotopez. Although I can’t say I understand all of this, would the main difference in speed between the console and CPS3 versions be due to the difference in refresh rates, or would you put it down to how 3rd Strike is emulated in OE?
Some simple maths shows that, if we assume the refresh rates are the only difference, CPS3 3S runs at ([59.583*100]/60 =) 99.305 % the speed of OE. If only there was a way to hack the game to slow it down by this amount…
I’ve tried running shmupmame at 0.99305 speed but it still feels too fast, but I haven’t tried any in-depth tests.
the difference in timing becomes more apparent the longer you need to keep strict timing.
think of it like adding 1 to the timing of each elbow. so because you added 1 on the first, your second elbow is already mistimed slightly (1) and then adding another 1, so now your third is off by 2, etc.
so a few times in you’re off by a wide enough margin that it won’t work with the same input timing as it did on the other setup.
I think it means the perceived difference in gameplay speed is just down to refresh rate. I expected going in that we’d find a smoking gun like “you can cancel Chun’s low forward into super on CPS3 much later than on OE” since it definitely feels that way, but it didn’t seem so. you can tell from how the tight links dropped on OE that there’s definitely a timing difference, and yeah for anything that requires tight timing several times in a row it’ll eventually drop.
so I think it’s fair to say what people have always said - OE is not the same as CPS3 but it’s reasonably close. if you play characters that require tight links you’ll probably notice the difference. if you don’t, you’ll probably have a slight bit more time to react to things on CPS3 but it’s not huge.
we haven’t tested audio latency at all, and that could be a factor too. if OE had a longer audio buffer you could be hearing your hit sounds later and would have less time to confirm if you confirm off sound. it’s definitely possible. any stage or move that causes slowdown on CPS3 will make a difference as well. and of course if you’re playing on a laggy stick or monitor or playing online, things will be really different (goes without saying).
also regardless of our results, OE’s sound effects still suck. there’s no cure for Oro snoring and trashcan shoryus.
The games do run at different speeds, so if you’re a robot with a super accurate ability to push buttons at specific times, you might notice the difference in speed if you try doing long combos like elbow cannons, tth on the twins, or some very timing specific genei jin combos. Whatever combo you’re doing though has to last long enough for your inputs to desync with the game speed, like tebbo was saying. That of course only matters if you’re used to playing on 1 platform and you switch to the other. Most people probably can’t do x6 elbow cannons on urien consistently without double tapping buttons. If they can though, they have visual and audio feedback that they probably depend on more so than an internal clock in their head. I can’t say that is true about everyone, but for most people it is.
I think there may also be a perceived difference in timing just based on how easy it is to get your input to come out on the frame you intended it to come out on between cps3 and oe. Perhaps USB on 360 is prone to occasional lag, which would change the timing of your inputs. This would only become apparent if you press your button on the very edge of cancel or link windows i.e if the move comes out 1 frame late you drop your combo.
“The games do run at different speeds”
"'May be a a perceived difference in timing"
If the game runs at a different speed then it’s not a perceived difference in timing but actually a difference in timing
I added an “also” to my post (May also be a perceived…). I see them as 2 different reasons people will say there’s a timing difference. The actual timing difference is very minute whereas a perceived difference isn’t really a timing difference, but it’ll feel like one and it would probably be more noticeable.
that plus any input lag differences make it easy to notice/feel.
it’s possible to sense something which measurements might tell us is essentially negligible. humans are remarkably acute.
Well… A lot of audiophiles claim they can “sense” differences between music files/sound systems that they really can’t (i.e. the difference between a 320kbps MP3 vs. lossless audio), which lead them to spending thousands on equipment that don’t really make a difference.
But that’s perceiving differences in sound quality, while we’re looking for differences in speed. Humans might be a bit better at that.
so you’re saying without knowing already that we couldn’t tell the difference?
we might exaggerate sure. but the game feels different. 360 OE on a crt is pretty good though.
not everyone might notice, but some probably do 100% of the time. there is 3S and then there are lesser versions of 3S.
even if its small, differences exist and they are noticeable. they do add up. it’s immediately apparent in my experience. super links are an easy example.
so like isotopez is saying. it might be pretty small, but the actual effect of say 1 frame of input lag more, can be pretty enormous (perceived). especially if the gameplay is running faster and the inputs are running slower (like the ports). so in the end you just have that much less time to react to something. teching throws is easier on cps3, period.
sharp in contrast deez nuts