MAME shmups input delay list
MAME shmups input delay list
After playing around with shmups in mame and in ports/pcbs for some time, it becomes harder to enjoy shmups/games which have "high" input delay while playing in Mame.
So I decided to make a list of the games that should be "avoided" or that do not have "optimal" response in mame.
Keep in mind that the list only shows the Mame delay and not the delay coming from other parts of the system (devices, usb protocol, windows, game controllers drivers). The delay was calculated using the mame skip to next frame button.
Also, this does not mean that the problems come solely from mame, depending on the way the games were coded and how they operated on the original hardware Mame can simply be reproducing a delay that the PCB actually has.
Updtade (March 20 2010) : Shmupmame Lagless Edition
-A mame version made by yours truly that reduces the input delay on most of these games.
So I decided to make a list of the games that should be "avoided" or that do not have "optimal" response in mame.
Keep in mind that the list only shows the Mame delay and not the delay coming from other parts of the system (devices, usb protocol, windows, game controllers drivers). The delay was calculated using the mame skip to next frame button.
Also, this does not mean that the problems come solely from mame, depending on the way the games were coded and how they operated on the original hardware Mame can simply be reproducing a delay that the PCB actually has.
Updtade (March 20 2010) : Shmupmame Lagless Edition
-A mame version made by yours truly that reduces the input delay on most of these games.
Last edited by nimitz on Sat Mar 20, 2010 1:45 pm, edited 4 times in total.
The list
The list was made using WOLF 0.128, new WOLF versions (over .120 or so) have input delay fixes.
Notes:
The lowest delay is always shown, for example if a game has 4 frames in WOLF99 and 3 frames in WOLF128 the game will be put in the "3 frames" category.
The actual "delay " is one frame less than specified in these categories, I decided to put it this way because this is how people usually refer to it.
All psx based games seem to have at least 3 frames delay or more (Sonic wings limited and Psyvariar have 4). This includes the newly emulated Taito G-net.
**These raizing games (Batrider, Bakraid) do not have "constant" delay, afaik the pcbs have 3 frames delay and mame seems to be adding an additional frame of its own(the mame induced part isnt constant though, seems to vary between 0 and 1 additional frame). Geragga exhibits the same behavior if you are not using Garemame (always 3 frames delay in garemame). Mahou seems to be the same only 2/3 frames.
Code: Select all
7 FRAMES (ludicrous)
Success:
Guardian Force
6 FRAMES (horrendous)
Namco:
Pistol Daimyo no Bouken
5 FRAMES (horrible)
Namco:
Galaga '88
Phelios
Dragon Saber
Dangerous Seed
Cosmo Gang - The Video
4 FRAMES (bad)
Psikyo(These 4 games have 3 frames in Raine):
Sengoku Ace
Gunbird
Strikers 1945
Tengai
Namco:
Tenkomori Shooting
Dragon Spirit
Blast Off
Others:
Mars Matrix (Takumi)
Radiant Silvergun (Treasure)
Sonic Wings Limited (Video System)
Psyvariar/Psyvariar Revision (Success)
3 FRAMES (passable)
Cave:
ESP Ra.De.
Guwange
Progear
Raizing:
Mahou Daisakusen (2.5 frames)**
Armed Police Batrider(3.5 frames)**
Battle Bakraid(3.5frames)**
Battle Garegga (3 frames using garemame, otherwise 3.5)
Soukyugurentai
Dimahoo
Shippu Mahou Daisakusen
1944 (WOLF99)
Namco:
Namco Classics Collection Vol.1 (Galaga, Xevious)
Xevious
Galaga
Sky Kid
F/A
Nebulas Ray
Ordyne
Capcom:
Area 88/ U.N. Squadron
U.S. Navy/Carrier Air Wing
Varth
19XX
1941
Eco Fighters
Forgotten Worlds
Taito:
Rayforce
Darius Gaiden
Grid Seeker
Gekirindan
Ryu Jin
S.R.D. Mission
Exzisus
Toaplan:
Tiger Heli
Slap Fight
Outzone
Truxton
Zero Wing
Hellfire
SNK:
Prehistoric Isle in 1930
Alpha Mission
HAL21
Banpresto:
Super Spacefortress Macross
Super Spacefortress Macross II
Macross Plus
Irem:
X-multiply
R-type Leo
Others:
Repulse (Sega)
Thunder Force AC (Sega)
Thunder Dragon 2 (NMK, 4 frames in WOLF99)
GunNail (NMK)
Sienryu (Warashi)
Strikers 1945 Plus (Psikyo)
Twinbee (Konami)
Nostradamus (Face)
Gyrodine (Crux/Toaplan*)
Chimera Beast (Jaleco)
Cybattler (Jaleco)
Cyvern (Kanko)
Sengeki Striker (Kaneko)
Aero Fighters (Video system)
Daioh (Athena)
Blazing Star (Yumekobo)
Blaze On (Atlus)
Boogie Wings (Data East)
R-Shark (Dooyoong)
Gondomania (Data East)
Sol Divide (Psikyo)
PSX based games(3 frames):
G-Darius (Taito)
Raystrom (Taito)
Ray Crisis (Taito)
Brave Blade (Raizing)
XII Stag (Traingle Service)
2 FRAMES
Most games (if a game is not elsewhere it's probably there)
1 FRAME(no delay, action happens on the next frame)
Taito:
Space Invaders
Darius
Halley's Comet
Metal Black
Asuka & Asuka
Gun Frontier
Konami:
Gradius
Gradius III
Flak Attack/MX5000
Juno First
Ajax
Nichibutsu:
Moon Cresta
Terra Cresta
Dangar UFO robot
Capcom:
1943 / 1943 Kai
Gun.Smoke
Others:
Hyper Duel (TechnoSoft)
Varia Metal (Excellent System)
Bakuretsu Breaker (Kaneko)
Change Air Blade (Sammy)
Storm Blade (Visco)
Satan's Hollow (Bally Midway)
The lowest delay is always shown, for example if a game has 4 frames in WOLF99 and 3 frames in WOLF128 the game will be put in the "3 frames" category.
The actual "delay " is one frame less than specified in these categories, I decided to put it this way because this is how people usually refer to it.
All psx based games seem to have at least 3 frames delay or more (Sonic wings limited and Psyvariar have 4). This includes the newly emulated Taito G-net.
**These raizing games (Batrider, Bakraid) do not have "constant" delay, afaik the pcbs have 3 frames delay and mame seems to be adding an additional frame of its own(the mame induced part isnt constant though, seems to vary between 0 and 1 additional frame). Geragga exhibits the same behavior if you are not using Garemame (always 3 frames delay in garemame). Mahou seems to be the same only 2/3 frames.
Last edited by nimitz on Sat Mar 20, 2010 12:46 am, edited 18 times in total.
-
powerfuran
- Posts: 177
- Joined: Wed May 03, 2006 2:00 pm
- Location: Italy
- Contact:
-
freddiebamboo
- Posts: 1366
- Joined: Sun Jan 30, 2005 9:17 pm
- Location: UK
I've never experienced any issues with batrider, bakraid or sengoku ace using wolf 0.99.
Strikers 1 does have noticable lag however for me, also in 0.99. Esprade is the only cave I've ever seen lag in, but that was several mame builds ago. This is just from playing obviously and not crazy frame input tests.
Strikers 1 does have noticable lag however for me, also in 0.99. Esprade is the only cave I've ever seen lag in, but that was several mame builds ago. This is just from playing obviously and not crazy frame input tests.
Doesn't make any difference afaik. (also, some of the results have been confirmed by others)bcass wrote:What are your hardware specs?
freddie:
Wolfmame .99 is indeed a pretty decent version concerning input lag, most of the games are on par with new WOLF command line version (like the one I used) the RF series is are good examples that differ (3 frames delay in wolf99). But I can assure you Gunbird, Strikers 1 and Sengoku Ace all have the same delay problem in EVERY mame version (afaik).
also, Batrider is a weird one, sometimes the input goes down to 3 frames using wolf99 or wolf101 and seems to stay at 3 frames. Bakraid is probably similar. 3 frames can be "hard" to notice depending on your ship speed and what kind of shmup your playing.
edit : I modified the OP to clarify the delay concerning Raizing games.
Nice list. Also, Storm Blade is misspelled.
| My games - http://www.emphatic.se
RegalSin wrote:Street Fighters. We need to aviod them when we activate time accellerator.
I'm affraid that it does matter. I've used several USB adaptors over the years, and some have introduced significantly more lag than others. Producing a list is all very well, but without a full technical specification of all the hardware used, it is useless. A scientist wouldn't publish results of an experiment without listing all his apparatus (and configuration of said apparatus).nimitz wrote:Doesn't make any difference afaik. (also, some of the results have been confirmed by others)bcass wrote:What are your hardware specs?
-
GaijinPunch
- Posts: 15721
- Joined: Mon Jan 31, 2005 11:22 pm
- Location: San Fransicso
I will post some from my own system (keyboard AND USB), but first, it's worth noting that the number of frames 'till the input catches can be variable. When I last experimented I noticed that sometimes I'd get, say, three frames, and other times four.
How exactly did you count? I would press hard right until the ship started moving in that direction, then left, and start counting the frames until the ship started moving right again. It might also be fun to see if the time it takes to re-center is any different.
How exactly did you count? I would press hard right until the ship started moving in that direction, then left, and start counting the frames until the ship started moving right again. It might also be fun to see if the time it takes to re-center is any different.
bcass : the test is not real time (like explained in the first post) so the other variables that you are mentionning are not present. Also i did test using different usb devices and keyboard and all produce the same results. I'll repeat one last time, this is the Mame delay and this should be the same for every system if you are using the same technique and WOLF Mame 0.128.
(Anyone who wants to recheck any of the games on the list feel free to do it and post your results here)
Ed : Nice to know you will try some on your system, can you please use this version (or WOLF 126/131)
http://mahlemiut.marpirc.net/
also, I counted the number of frame it took to respond INCLUDING the "input" frame. For example : Hold left, skip, skip, *Moves*. would be "2 frames"
also, very few games have variable input using new mame versions with "fixed" input (like newer WOLFmame). Raizing games come to mind, if you find anything else please tell us.
(Anyone who wants to recheck any of the games on the list feel free to do it and post your results here)
Ed : Nice to know you will try some on your system, can you please use this version (or WOLF 126/131)
http://mahlemiut.marpirc.net/
also, I counted the number of frame it took to respond INCLUDING the "input" frame. For example : Hold left, skip, skip, *Moves*. would be "2 frames"
also, very few games have variable input using new mame versions with "fixed" input (like newer WOLFmame). Raizing games come to mind, if you find anything else please tell us.
Question number one: how many PCBs do you own?nimitz wrote:bcass : the test is not real time (like explained in the first post) so the other variables that you are mentionning are not present. Also i did test using different usb devices and keyboard and all produce the same results. I'll repeat one last time, this is the Mame delay and this should be the same for every system if you are using the same technique and WOLF Mame 0.128.
Question number two: what is your hardware specification?
Question number three: what is your hardware specification?
Question number four: do you understand the nature of the term 'real-time'?
Question number one: 27 bazillions
Question number two: 486 DX2 66
Question number three: windows 3.1
Question number four: True real time is impossible since the refresh rate of the screen is around 60hz, but when the action happens on the next frame when removing other factors from the equation, mame is doing its "job" of making the game react on the next displayed frame
ps. Can you please bring it to PM or come to #shmups on efnet if you want to continue this "discussion".
Question number two: 486 DX2 66
Question number three: windows 3.1
Question number four: True real time is impossible since the refresh rate of the screen is around 60hz, but when the action happens on the next frame when removing other factors from the equation, mame is doing its "job" of making the game react on the next displayed frame
ps. Can you please bring it to PM or come to #shmups on efnet if you want to continue this "discussion".
Ed Oscuro wrote:I will post some from my own system (keyboard AND USB), but first, it's worth noting that the number of frames 'till the input catches can be variable. When I last experimented I noticed that sometimes I'd get, say, three frames, and other times four.
Have a guess. Go on, have a guess why it's variable.Dandy J wrote:lol @ ppl asking about hardware differences
Anyhow, I'll have to check this out later, but I'll make sure to check out that version of MAME you want me to. lolnimitz wrote:Question number four: True low rez is impossible
Anyway, bcass, I dunno if this was noticed I was getting that variation in the same session, from one group of frames to the next. Wouldn't be terribly surprising if this just a hiccup moving from the Windows input buffer / whatever into MAME and the actual game's input system.
-
GaijinPunch
- Posts: 15721
- Joined: Mon Jan 31, 2005 11:22 pm
- Location: San Fransicso
I don't think the note under "1 frame" is strong/general enough. That phenomenon potentially affects all of the measurements, not just the difference between "1 frame" and "2 frames". Also, it's not just a matter of when MAME samples the input vs. when it pauses, but also of when the game itself samples its input. This might explain the difference between Mars Matrix and the other CPS2 games; also the difference between Pistol Daimyo no Bouken and the other Namco System 1 games.
Ex-Cyber : Very good point, I was thinking about that earlier and I'm pretty sure you are right.
I did some more testing and the games under "1 frame" and they really do seem more responsive than some of the "2 frames" ones.
My guess is, like you pretty much said, that the input for these game can be taken from RIGHT before the drawing of a new frame and still be updated. While the games with "2 frames" are handled differently and take a bit more time to be processed .
This leads me to believe that the delay is never really the same for two different systems, (i.e. it could be 1.5frames, 1.6 frames, 1.8 frames). So these categories do not actually tell us where a game is EXACTLY but is good to have a general idea of the delay.
GP : Pre- 0.106 vanilla version of MAME will yield very similar results (except on games with specific fixes, like RF series). That being said, using vanilla MAME after 0.106 (107 up) will give you much worse results, things like 6 frames delay with Batrider...
I did some more testing and the games under "1 frame" and they really do seem more responsive than some of the "2 frames" ones.
My guess is, like you pretty much said, that the input for these game can be taken from RIGHT before the drawing of a new frame and still be updated. While the games with "2 frames" are handled differently and take a bit more time to be processed .
This leads me to believe that the delay is never really the same for two different systems, (i.e. it could be 1.5frames, 1.6 frames, 1.8 frames). So these categories do not actually tell us where a game is EXACTLY but is good to have a general idea of the delay.
GP : Pre- 0.106 vanilla version of MAME will yield very similar results (except on games with specific fixes, like RF series). That being said, using vanilla MAME after 0.106 (107 up) will give you much worse results, things like 6 frames delay with Batrider...
And all of these "tests" are done @ 60kHz, right? Because that's a pretty important factor, considering far from all emulated games originally run @ 60kHz natively. Or has this already been taken into consideration?
| My games - http://www.emphatic.se
RegalSin wrote:Street Fighters. We need to aviod them when we activate time accellerator.
Well, in actual play a 1-frame variation in lag is unavoidable with a conventional raster display. For any given button/joystick activation you could close the switch right before or right after input is sampled for the next frame, just as MAME could be pausing right before or right after the input is sampled.nimitz wrote:This leads me to believe that the delay is never really the same for two different systems, (i.e. it could be 1.5frames, 1.6 frames, 1.8 frames). So these categories do not actually tell us where a game is EXACTLY but is good to have a general idea of the delay.
I guess you mean 60Hz and not 60KHz?emphatic wrote:And all of these "tests" are done @ 60kHz, right? Because that's a pretty important factor, considering far from all emulated games originally run @ 60kHz natively. Or has this already been taken into consideration?
First of all i think many of you guys don't understand the nature of the test, I am not testing the games during actual play, but I am using a feature of MAME that allows to skip to the next frame while the game IS PAUSED.
So, yes i am aware that some games run at 58Hz or even at 55Hz, but even at 55Hz the difference in absolute response time tough, is not high enough to cause the games to be classified under different frame categories.
These tests are very interesting, but, as you wrote yourself, the actual games should be tested, too. If programmed like crap, some might have several frames of native input lag, who knows ?
(it could be tested using the real thing, or by checking within MAME when games are notified of the user input - that is to say, not at the player level.)
(it could be tested using the real thing, or by checking within MAME when games are notified of the user input - that is to say, not at the player level.)
Last edited by Keade on Sun May 31, 2009 9:44 pm, edited 4 times in total.
Your guess is good.nimitz wrote:I guess you mean 60Hz and not 60KHz?
I've been lead to believe that this is major reason for lag, that's why I brought it up. Something about MAME not passing inputs to the emulation until the current frame has caught up with the monitor refresh rate. That's why playing near 50Hz games is better at 100hz (as that's dividable by an even number).nimitz wrote: So, yes i am aware that some games run at 58Hz or even at 55Hz, but even at 55Hz the difference in absolute response time tough, is not high enough to cause the games to be classified under different frame categories.
This would render this test useless though.
| My games - http://www.emphatic.se
RegalSin wrote:Street Fighters. We need to aviod them when we activate time accellerator.
This seems especially likely with some of the newer games using framebuffer-based hardware (such as PSX-based hardware or ST-V), since at least a few of them are probably double-buffered.Keade wrote:These tests are very interesting, but, as you wrote yourself, the actual games should be tested, too. If programmed like crap, some might have several frames of native input lag, who knows ?
Speaking of PSX-based hardware,
Sol Divide's hardware is not PSX-based.nimitz wrote: PSX based games(3 frames):
Sol Divide (Psikyo)
Indeed, ST-V games give the same sort of "unilateral" delay, which never is under "3 frames". I would be curious to know if PSX or ST-V emulators give that same kind of "base" or "emulated" delay.Ex-Cyber wrote:This seems especially likely with some of the newer games using framebuffer-based hardware (such as PSX-based hardware or ST-V), since at least a few of them are probably double-buffered.Keade wrote:These tests are very interesting, but, as you wrote yourself, the actual games should be tested, too. If programmed like crap, some might have several frames of native input lag, who knows ?
Thanks, for some reason I thought it was, probably had to do with the ugly pre-rendered graphics (:lol:) and the psx port.Ex-Cyber wrote:Sol Divide's hardware is not PSX-based.
Updated the list with some more games.
Found a shmup with 7 frames! (8 frames in wolf99). Guardian Force... if you want to see extreme delay try this out in WOLF99.
Also, after some more testing I decided put Batrider and Bakraid under "3 frames" since they seem to only have 4 frames delay under specific cir circumstances (right after boot up for example).
edit : after testing all the Namco shmups it seems only 2 of them have 2 frames delay : Galaxian and Gaplus (aka Galaga 3). The rest are between 3 and 6 frames delay.
Found a shmup with 7 frames! (8 frames in wolf99). Guardian Force... if you want to see extreme delay try this out in WOLF99.
Also, after some more testing I decided put Batrider and Bakraid under "3 frames" since they seem to only have 4 frames delay under specific cir circumstances (right after boot up for example).
edit : after testing all the Namco shmups it seems only 2 of them have 2 frames delay : Galaxian and Gaplus (aka Galaga 3). The rest are between 3 and 6 frames delay.
this is what i thought as well.....although i still think most of peoples complaints about input lag with mame is due to shitty adaptersemphatic wrote:Your guess is good.nimitz wrote:I guess you mean 60Hz and not 60KHz?
I've been lead to believe that this is major reason for lag, that's why I brought it up. Something about MAME not passing inputs to the emulation until the current frame has caught up with the monitor refresh rate. That's why playing near 50Hz games is better at 100hz (as that's dividable by an even number).nimitz wrote: So, yes i am aware that some games run at 58Hz or even at 55Hz, but even at 55Hz the difference in absolute response time tough, is not high enough to cause the games to be classified under different frame categories.
This would render this test useless though.