I can’t like this thread enough.
I thought MAME didn’t use the video card, all software rendering? So if that’s the case your video card isn’t the bottleneck.
Thanks for bringing this to our attention bro. I just had a quick round and it instantly felt better than the FBA GGPO uses. Night and day difference IMO
Shmup Mame is pretty cool. I use it for all my shmupping needs. Should be great for fighters as well.
Some good news. I did two tests tonight.
First using supergun -> crt, US Turbo 2
The crouching RH comes out majority of the time at 4 frames, with a few 3 and 5.
I would say the average number is 4
Raw footage here
The reason why the numbers are inconsistent is because of frame skipping
The 2nd test I did was crappy core2duo laptop with mobile 4 series express graphics output to CRT at 800x600 resolution. using shmupmame with 99% speed setting (everything else is stock) on windows 7 with aero disabled and mame running at full screen.
The crouching RH also comes out majority of the time at 4 frames, with a few 5.
I would say the average number is 4.3
I’ll do more test this weekend.
I plan to do
desktop i5 2500k with radeon 6800 => CRT
desktop i5 2500k with radeon 6800 => laggy dell lcd
2012 macbook => macbook lcd screen
core2duo laptop with mobile 4 series express graphics => laptop lcd screen
will keep you guys posted
Good stuff. Can you describe how you are getting the output from your 6800 to your CRT? Also, what kind of CRT are you using?
What we need to do now is what DGV and I talked about in a thread on here one time, and make a PS/2 Keyboard Arcade stick to avoid USB input lag.
soooo for us less informed, why would you use this particular emulator over regular mame? you can adjust the emulation speed in it too, in fact in the slider controls the main cpu is already at 74% by default, the significance of which i don’t know but if i crank it up to 100% the game speeds up quite a bit.
This version of MAME was designed by the shmup community for the sole purpose of reducing input delay. So for all intents and purposes, their version of mame works perfectly with what fighting game enthusiasts look for in an emulator as well.
The readme explains it best…
updated second post with a modified version of shmupmame with easier input mapping
Man, I headed down this path a little while back. Its not difficult to do, from a building perspective…BUT its difficult to locate a modern motherboard that has a ps/2 port that isn’t tied to its USB controller.
yeah. im waiting for undamn’s a board…
yeah. im actually stockpiling CRTs lol.
but really most lagless non CRT displays are really computer monitors. and I think there are plenty of Lagless Analog (RGB)-> VGA converters out there to use with ST boards…
I gave up on the vaporware known as undamned’s board. I use a MAK Strike supergun (small and cheap) running RGB into a Sony PVM-20L5 CRT. Super Turbo looks amazing on it…and of course there is 0 lag.
got more test results tonight.
shmupmame + 2012 macbook air => mini-displayport-to-vga => CRT at 800x600
same result as core2duo laptop => CRT
just to verify the claim of shmupmame having 1 frame less lag, I did a control test of
mame 0148 + 2012 macbook air => mini-displayport-to-vga => CRT at 800x600
mame has consistently 1 more frame than shmupmame
the last test is interesting
shmupmame + 2012 macbook air => macbook lcd at 1440 x 900 native resolution
half of them were 4 frames and half of them were 5 frames
average 4.5 which is extremely good.
not to mention portable!
(we played a bunch at NCR using this setup)
OK I think I’ve done all the meaningful tests I could think of
ps2 hyper street fighting turbo2 fei long stage timer
52.1 seconds (1 second faster than arcade)
ps2 hyper street fighting turbo2 => CRT (NKI test)
core2duo laptop shmupmame => laptop lcd (NKI test)
4.5 frames (same as macbook)
core2duo laptop ggpo final burn alpha us_turbo2 CRT (both fullscreen & window)
desktop i5 2500k with radeon 6800 => laggy dell lcd
Updated #3 post with all the results.
Thanks for the expansive testing! Can you tell us more about your setup post-computer? I.e., components, cables, CRT, etc.?
You can find out here.
One more update regarding playing laggy online matches using laggy monitor with a laggy emulator.
As shown above, final burn alpha has 2 more frames of input lag when playing offline on a lagless CRT.
2/60 = you are deprived of 3.33% of a second to react to any moves.
Stage time for ggpo final burn alpha turbo 2/1/0
turbo 2 (52.5 seconds)
turbo 1 (53.9 seconds)
turbo 0 (55.1 seconds)
With turbo speed set to turbo 1, the game is 2.67% slower overall. But it still does not feel as responsive as arcade.
With turbo speed set to turbo 0, the game is 4.95% slower overall, you have enough time to react but your input will still come out slower.
And that was for offline.
When online lag comes into play, you can see that playing at a slower turbo setting will be better.
HUGE thanks to Papasi for conducting all of these tests. This could definitely be an ideal solution for all those who wish to organize an ST event locally or at a major but do not have access to arcade equipment. All that’s really needed is further playtesting. This is where the ST community comes to play. If we can get some thorough playtesting amongst the entire community we can easily come to find the best settings that match the arcade pcb identically. Case in point, Papasi mentioned at 99% Speed on T2 (US version) seems to be the best to mirror the US board, whereas on 97% speed on T3 (JPN version) seems to mirror the JPN board on my end.
At present based off Pap’s test, ShmupMame is REALLY close to matching the PCB in terms of input lag. It’s actually even more responsive than the DC version. So what that actually means in terms of practical use --> Shmupmame 4.2 + Evo Asus monitor = roughly around 1 frame above Arcade, which is exactly the DC version (DC = +1 frame compared to Arcade). This already places the setup well ahead of Classic Mode (360) and even better than DC, since that version is actually constrained to a CRT, not unless one uses the VGA Box for DC (which looks extremely pixelated). Of course, we as the community can strive for even better than just settling with 1 frame (which is really good btw!), by opting to use a better LCD than the Evo Asus(there are quite a few listed on the sub-1frame monitor thread via Tech Talk). There’s even the possibility that the emulator can actually match the board 100% by just adjusting more settings within the Emulator…we just need more people to actually test and compare notes =)
The only problem with the PS/2 Keyboard route is that it is limits who can utilize their own hardware. Sure the organizer can have a few modified keyboard arcade sticks, but that does not solve the issue of allowing players to use their own stick/pad and in essence, forces a particular controller on the players. This method is also what limited the supergun route in terms of flexibility, until Undamned changed everything by releasing his consolised CPS2 setup and usb jamma adapters.
Regarding USB input lag…has it been 100% confirmed that all arcade usb sticks actually poll at 125 hz (8ms) on a pc? I did the USB poll rate hack (which for some reason only recognizes my MadCatz SFxTekken stick, but not my modded sticks via PS360 chip) and I didn’t notice any difference between playing at 1000hz (1ms) or the default 125hz. Which makes me wonder if either:
A) all arcade sticks poll at 1000hz
B) the USB poll rate program didn’t actually register the stick