CAVE SLOWDOWN: Ports Vs. The Real Thing

segasonicfan · Post by **segasonicfan** » Tue Aug 01, 2006 8:46 am

I got a simple question for you Cave PCB owners out there: how does the slowdown on PS2 ports compare with the real thing? I was playing DDP-DOJ, Ibara, and Espgaluda on my PS2 today and there were some spots of major slowdown...all on 1 player. DOJ had the most, almost completely stopping bullet movements at some points. From my experience with my Esprade and Dangun boards, 1Player gaming is virtually slowdown-free. Is it the same with other titles? That would explain some of Mushi's PCB value as well....

One of the reasons I ask BTW is I am interested in buying Ibara/DOJ PCB's and it would be nice to know the differences.

-Segasonicfan

NTSC-J · Post by **NTSC-J** » Tue Aug 01, 2006 8:54 am

Not sure where it is in DOJ that has slowdown so bad it freezes the action, but the PCB has more overall. ESPGaluda is a 95% perfect port, so whatever slowdown that's in the PS2 game is likely in the PCB as well. Ibara on PCB has way more slowdown than the PS2 port, the train in stage 3 being the biggest offender.

CIT · Post by **CIT** » Tue Aug 01, 2006 9:11 am

Actually a lot of people were complaining that Mushihimesama was missing slowdown on the PS2 port, thus making some sections more difficult.

EOJ · Post by **EOJ** » Tue Aug 01, 2006 9:21 am

Mushi is missing bits of slowdown, mainly in parts of Stage 3 and the Stage 5 midboss. But it's about 90% accurate overall (at least, when I play the PCB, I don't notice any differences other than the parts I mentioned). However, the slowdown is a bit more jerky in the PS2 version (it slows down then the speed rockets up again a bit too fast) in some parts and many people complain about this. It doesn't bother me though. It's a much better port in terms of slowdown emulation compared to Ibara, as the Ibara port basically strips away all of the PCB's slowdown.

Plasmo · Post by **Plasmo** » Tue Aug 01, 2006 9:29 am

unfortunately most of the ps2 ports dont have the pcb slowdown

GaijinPunch · Post by **GaijinPunch** » Tue Aug 01, 2006 9:30 am

90% is WAAAAY too generous. Talking slowdown only, I give it an 80%, but that's only counting the places there are slowdown. Then you've got the jerky effect. This doesn't count the turd display resolution.

I think the standard grading of the recent games is:
DOJ - 90%
ESPGaluda - 99%
Mushi & Ibara < 80%

DOJ & Galuda are the only two (that I can think of ever) that go out of their way to emulate the slowdown.

For bonus, DDP, ESPRade, Dangun, and I think Guwange as well seem to be not only missing slowdown in MAME, might even be running a bit fast.

Post by **system11** » Tue Aug 01, 2006 9:39 am

Slowdown is hard to emulate, because it's inconsistent and unintentional unless you code in routines that forcibly drop frame update depending on number of bullets present. It's not even the same across identical PCBs sometimes.

True.

Post by **Dave_K.** » Tue Aug 01, 2006 2:31 pm

bloodflowers wrote:Slowdown is hard to emulate, because it's inconsistent and unintentional unless you code in routines that forcibly drop frame update depending on number of bullets present. It's not even the same across identical PCBs sometimes.

True.

Unfortunally, this is very true. I bought a second DDP board just to compare and there was a noticeable difference between the two. Even side by side when started at the same time, they would desync before geting around to displaying the cave logo again. Believe it or not, but the second board appears to be just as fast as Mame

.

To be more on topic, I feel the ports are close enough that any average player should not be concerned with the accuracy of emulated slowdown, and instead concentrate on the scoring. If you start approaching the top of the high score threads here, then I'd think about getting serious and using the PCBs to get a more accurate score.

LUNardei · Post by **LUNardei** » Wed Aug 02, 2006 12:52 am

Tried ULTRA on PCB the other week: sad to say, in some moments the PCB is WAY faster.

Valgar · Post by **Valgar** » Wed Aug 02, 2006 12:56 am

Ibara PS2 has like no slowdown.

EOJ · Post by **EOJ** » Wed Aug 02, 2006 2:26 am

GaijinPunch wrote:Mushi & Ibara < 80%

Lumping the Mushi and Ibara ports together gives the wrong impression. Mushi is 80-90% accurate, while Ibara is 0% accurate, in terms of slowdown.

segasonicfan · Post by **segasonicfan** » Wed Aug 02, 2006 3:14 am

I think you guys are missing what I was saying. I made the poor assumption that the PCB's had LESS slowdown...but it seems the other way around. I'm not trying to emulate slowdown for scoring purposes or anything like that, I'm trying to get rid of it completely

Weird, I just figured the actual PCB's would feature less of it being that they were designed specifically by Cave.

-Segasonicfan

EOJ · Post by **EOJ** » Wed Aug 02, 2006 3:35 am

segasonicfan wrote: I'm not trying to emulate slowdown for scoring purposes or anything like that, I'm trying to get rid of it completely

Well, then you should DEFINITELY buy the Ibara PS2 port. You'll love it. The game is smoother without slowdown, admittedly, and a bit more of a challenge. It seems some companies like to remove slowdown when porting shooters (perhaps this is what Cave did in Ibara?), they think slowdown is a negative byproduct of the PCB hardware. A good example is Treasure, who are on record for saying the GC version of Ikaruga is "better than the arcade" because it does away with all the Naomi version's slowdown, something not possible on the Naomi hardware itself due to its limitations.

I prefer the PS2 version of Mushi, primarily for the rapid fire settings, but also somewhat because sometimes the PCB slows down at points where the action on the screen doesn't really need any slowdown to help you out. Feels like you're playing underwater.

BeauZilla 002 · Post by **BeauZilla 002** » Wed Aug 02, 2006 4:12 am

Unfortunally, this is very true. I bought a second DDP board just to compare

Damn,can I come work with you?

GaijinPunch · Post by **GaijinPunch** » Wed Aug 02, 2006 1:17 pm

Lumping the Mushi and Ibara ports together gives the wrong impression. Mushi is 80-90% accurate, while Ibara is 0% accurate, in terms of slowdown.

That would make DOJ and ESPGaluda over 100%, which is universally impossible (last I checked, anyway).

ill6 · Post by **ill6** » Wed Aug 02, 2006 1:24 pm

All you should be worried about is the speed of the original. If the original has slowdown you want the slowdown because the games are designed around the slowdown. If it goes faster in these places then you are gonna die more on a port than on the PCB.

Personally I find playing the games on PCB much easier than on a port, maybe because of slowdown but I think the main reason is that you sit at the cab in the right angle for the game. The game was designed and tested at a sitdown candy and that works best for me.

EOJ · Post by **EOJ** » Wed Aug 02, 2006 6:25 pm

GaijinPunch wrote:
Lumping the Mushi and Ibara ports together gives the wrong impression. Mushi is 80-90% accurate, while Ibara is 0% accurate, in terms of slowdown.
That would make DOJ and ESPGaluda over 100%, which is universally impossible (last I checked, anyway).

How does Ibara being 0% accurate in slowdown emulation (as there is no slowdown!) make DOJ and Espgaluda over 100%? I don't get the logic there.

tekneekz · Post by **tekneekz** » Wed Aug 02, 2006 9:44 pm

lmao that worthy of owned?

GaijinPunch · Post by **GaijinPunch** » Thu Aug 03, 2006 5:38 am

How does Ibara being 0% accurate in slowdown emulation (as there is no slowdown!) make DOJ and Espgaluda over 100%? I don't get the logic there.

It doesn't, but saying Mushi is 80% does. Based off an imaginary scale of TAC saying DOJ was about 90% accurate (and that was with VERY painstaking efforst to emulate the slowdown), Mushi can't be 80%. The "jitter" alone knocks off most of the accuracy, and there's quite a few parts where the slowdown is just not there. Spots in stage 3, etc. Saying Mushi's is 80% accurate isn't fair to the other ports.

Again, you can't really measure it though, but I think Mushi's is pretty bad. Never played Ibara other than that one time you brought it over, so can't really comment.

EOJ · Post by **EOJ** » Thu Aug 03, 2006 6:06 am

Okay, I get what you're saying now. But you yourself gave mushi an 80% earlier in this thread, I was just quoting that estimation.

It's difficult to quantify the 'slowdown accuracy' factor, as you said. What we're tossing about in this thread are partly subjective estimations. The Mushi port DOES have loads of slowdown, in parts where you need it. Stage 5 is a nice example (sans the midboss), as it has it everywhere you need it (and it's not jerky there either, unlike in the previous stages). So because of that I think it's a great port. Not perfect by any means, but I've gotten a great deal of enjoyment out of it so that qualifies as 'great' for me. Like I said, I notice few differences when I play the PCB in terms of 'slowdown spots'. You've played the PCB way more than me though, but I've played the PS2 port way more than you (I'd guess so at least, you've never seemed too keen on it). So we're sorta coming from different perspectives.

Anyway, Ibara PS2 is a different story. No slowdown anywhere. When I played the PCB in Japan the game really felt different. I didn't get that same vibe when playing the Mushi PCB in Japan.

louisg · Post by **louisg** » Thu Aug 03, 2006 6:37 am

whyyyyyyyy

GaijinPunch · Post by **GaijinPunch** » Thu Aug 03, 2006 9:21 am

twe wrote:Okay, I get what you're saying now. But you yourself gave mushi an 80% earlier in this thread, I was just quoting that estimation.

It's probably easier if I say this. 80% accuracy to me sounds like a "good" job. Not great, but good. Obviously part of my view is tainted after the work that Arika did, but I just can't say that Mushi is a "good" port. "Fair" or "Playable" I can say pretty easily.

DEL · Post by **DEL** » Thu Aug 03, 2006 9:42 am

I can only refer to the Mush PCB and PS2 Port.
The PCB (Maniac) has slowdown in certain places that allows me to 'enjoy' the game. It gives small breathing spaces.

I hear that Ibara PCB (train) has similar slowdown.

Is this PCB slowdown programmed in on purpose? = Probably, and personally...I like it.

Slowdown for The Old Folks

Post by **system11** » Thu Aug 03, 2006 10:54 am

Slowdown only exists because the hardware can't cope. There are a few approaches to resolving it:

1) Improve the hardware (not going to happen)
2) Let it slow down
3) Work out how many bullets roughly you can move before uncontrolled slowdown hits, and shift the entire game engine to 30fps when something a little short of that number is reached. This smoothes things out a little.
4) Purposefully put in slowdown timings according to bullet spreads. This is what a few people seem to think is happening.

At best, Cave are using option 3, but looking at the games running, I've never personally believed that to be true, or at least not 100%, it still seems highly variable and frequently uncontrolled. For example, there are big spreads in Mushi which slow down a lot, if it was any kind of graceful timed slowdown, rather than essentially hardware driven, they would not speed back up before clearing - but they do.

I personally believe a lot of it is a combination of wishful thinking and a belief that Cave are perfect, coupled with an old and incredibly misleading comment from someone on the MAME Testers page, claiming that it's intentional just because the emulation isn't maxing out - which is an insane proposition.

Double Dragon slows down like a bitch without maxing emulation, the board does it too. This is not intentional, it's because the CPUs aren't fast enough, and MAME is emulating the same CPU speed. For bonus points, it's a trivial change to eliminate slowdown in Double Dragon - I'll leave that to the reader to work out how. I'm working on a hardware mod for the PCBs.

louisg · Post by **louisg** » Fri Aug 04, 2006 4:41 am

Yeah, generally (on the vast majority of emulators) a symptom of maxing out your real CPU on emulation is that it will drop frames and not actually slow down more.

AFAIK slowdown seems to be caused by the game missing a vblank and then having to wait for the next one to update. Maybe it could be smoothed by enforcing a minimum slowdown period (not immediately returning to full speed)?

zaphod · Post by **zaphod** » Fri Aug 04, 2006 5:34 am

Yeah most arcade games will sync to the vblank, an when you can't do everything in one vblank the frame gets doubles and you have slowdown.

Cave shooters tend to have a mix of intenional and unintentional slowdown. DOJ is a perfect example. There are parts where the PCB REALLY slows down and chugs, and this is never intended, but it happens. Ad ther are artifcal slowdown that tends to kick in a well predicted times. When it is known that slowdown is going to happen, the good companies will try and manage it so that it results in less cheap deaths. If, fro example R-Type didn't slow donw whe things really got hectic in stage 7, people would go "WTF??" There's a cetain point of action at which slowdown is almost expected, and in most games that reach that point, the slowdown is indeed present.

As for the MAME is too fast issue, this has to do with wait states. Even "identical" PCBs may have memory with different wait states on it, and this causes diffrerences in the slowdown. MAME has no wait states, hence less slowdown. With some games it's particularly noticable, like DAngun Feveron.

Wait state emulation is incredibly tricky, so most emulators justpretend they don't exist, and most games don't need the emulation to be that exact.

GaijinPunch · Post by **GaijinPunch** » Fri Aug 04, 2006 3:02 pm

After playing the PCB and PS2 port of Mushi, I think it's fairly obvious that either the SH3 hardware itself knows it's limitations, and gradually changes frame rates as it reaches these limitations. OR, it's programmed into each game, most likely done w/ similar routines. The PS2 version is an example of what happens when a shitload of bullets are just thrown at the PCB w/o trying to manage the slowdown at all.