SHMUP Slowdown / Accuracy of Arcade Cores

PikWik
Posts: 256
Joined: Sat May 30, 2020 7:00 pm
Has thanked: 192 times
Been thanked: 71 times

SHMUP Slowdown / Accuracy of Arcade Cores

Unread post by PikWik »

9HUdDg0.jpg
9HUdDg0.jpg (37.57 KiB) Viewed 9802 times

With the recent release of the 1st gen cave game, dodonpachi, it has opened a laundry list of questions from me.
shmup fans are categorically the most analytical about how arcade emulation functions and how a game reacts to their movements.
to my untrained eye, a shmup game "plays" as it should, but then i read how there are little things which are not quite like a real PCB.
which leads me to think "shouldnt these FPGA recreations be exactly as an arcade?" :?:

i think some of the problem is from the amount of people who have access to both the original PCBs and a MiSTer to compare the two.
with mainstream retro consoles, there is a much larger group of people that can compare actual hardware to the FPGA mister version, and for the most part, the popular consoles (SMS, NES, genesis, SNES, neogeo, turbografx, GB, GBA) are at a point where the hardware-to-FPGA difference in behaviour is ~99% accurate with very few accuracy issues in speed.

the fact that fewer people have original arcade PCBs and a mister to compare are, imo, what will ultimately hold back a core developer from really nailing down the behaviour of arcade games to that same ~99% accuracy of the mainstream retro consoles. im under the impression, an arcade core developer should be working close with someone who has a PCB, is exceptionaly good at the game, and understands every small piece of how the original PCB behaves and the slowdown for the game.

i bring up these FPGA accuracy concerns because MAME does exist. more notably groovymame.
with something like groovymame, which is hardware accurate in terms of slowdown for a large majority of their shmup games, the 1-frame of inherent lag from software emulation is a valid tradeoff, compared to playing an FPGA version of an arcade that is inaccurate. yes, it is cool to have all the arcade parts of the PCB FPGA'd, but thats not enough to get the FPGA core to a place where its considered 1:1 accurate to a PCB.

i do remember when CPS1 was being worked on, people were comparing the core to actual PCBs, but im not seeing this same dedication with arcade cores lately. my thought is, "if accuracy of the original hardware and preservation is the end goal of FPGA, then the majority of these arcade cores are works-in-progress and shouldnt be used for anything except educational purposes." groovymame, for all intents and purposes, can function the same if not better than their FPGA counterparts, and groovymame in particular can work to reduce the 1-frame of software lag down to single digit milliseconds and not 16ms.

this is in no way a slight against the MiSTer.
i love having these retro systems in FPGA form, and still view my MiSTer purchase as one of the coolest pieces of electronics ive bought in a very long time. but i am concerned with arcade core development as a whole, and feel like an arcade developer should be working in tandem with a person who knows the PCB and how it plays. if the general public plays an arcade core, im very sure they likely wont know the difference, but getting a consensus for arcade FPGA accuracy, in particular from SHMUP fans, is a much trickier and long winded process.
AntialiasedPixel
Posts: 44
Joined: Mon May 25, 2020 2:17 am
Been thanked: 2 times

Re: SHMUP Slowdown / Accuracy of Arcade Cores

Unread post by AntialiasedPixel »

I think you're underestimating how much the devs have played the real PCB. As I understand it, most of them have the PCB in their possession and have likely played the game on the original PCB hundreds of hours by the time the core is nearing completion.

I don't know how close these cores are at this time, but it's not like improvements aren't being made fairly consistently. Just because something isn't 100% accurate out the gate doesn't mean it is considered "done" and things that are noticed will get fixed if people want to put the time in. People that follow mister development all know that changes/improvements to accuracy are always being worked on, and I don't think many claim that the majority of cores are "perfect" yet.
XtraSmiley
Posts: 109
Joined: Wed May 27, 2020 11:33 pm
Has thanked: 101 times
Been thanked: 15 times

Re: SHMUP Slowdown / Accuracy of Arcade Cores

Unread post by XtraSmiley »

While I agree none of these cores seem 100%, these developers are awesome and many of them do indeed have the actual PCBs, so that shouldn't be a worry.

I know there are many people able to donate or lend PCBs too (myself included, PM me!) so devs can continue their awesome work.

Don't sweat it so much, enjoy the games and report bugs. That's all any of us can do, unless we know how to do FPGA programming of course! :)
grizzly
Posts: 375
Joined: Tue Jun 16, 2020 12:22 pm
Has thanked: 55 times
Been thanked: 76 times

Re: SHMUP Slowdown / Accuracy of Arcade Cores

Unread post by grizzly »

Some valid points from all three before me!
If we then say a mister core should be 100% accurate too the arcade board (or console/computer).
Then how do we then deal with the small difference on different boards (or consoles/computers) due too components tolerances, and that is after all aging components have been replaced already?

In the speedrunning world it´s not uncommon that they have bought/sold many consoles/etc to get one with the right amount of tolerances on the components so they can win a couple of milliseconds here and there.
I mean where in that spectrum should the mister core "control" board be?
Middle? How do we then know it is in the middle? the answer is we must try many,many boards!
User avatar
aberu
Core Developer
Posts: 1144
Joined: Tue Jun 09, 2020 8:34 pm
Location: Longmont, CO
Has thanked: 244 times
Been thanked: 388 times
Contact:

Re: SHMUP Slowdown / Accuracy of Arcade Cores

Unread post by aberu »

You really gotta put things in perspective here.

1. MAME as a project has been around for almost 24 years now. The MAME arcade drivers weren't as accurate as they are now the entire time :P

2. MAME, as a result of being an emulator in software that runs on PC's, has significant latency overhead, so the comparison is sorta apples to oranges. There may be some cores that will have temporarily some bugs, but the input lag of all software emulators should probably be counted into that comparison then. No emulators "feel" the same to me, and that's always been why it was hard for me to go through games entirely in emulators, and I've been using emulators since the mid-90's. Pretty much only modern-day versions of Mednafen with PS1 rpgs which don't require the tighter inputs have been the only thing that would feel okay.

3. These core developers often do have the boards as others have said, or they are allowed to borrow them from a collector, and I know for a fact that patreon money from many core developers (not necessarily just arcade cores) often go toward purchasing hardware and testing gear like complex logic analyzers, etc.. so they can make their core better.

I get the passion. Report the bugs when you see them, it's a long-term project.
birdybro~
zakk4223
Posts: 270
Joined: Sun May 24, 2020 10:55 pm
Been thanked: 107 times

Re: SHMUP Slowdown / Accuracy of Arcade Cores

Unread post by zakk4223 »

AntialiasedPixel wrote: Sat Jan 16, 2021 11:35 pm I think you're underestimating how much the devs have played the real PCB. As I understand it, most of them have the PCB in their possession and have likely played the game on the original PCB hundreds of hours by the time the core is nearing completion.
(Not commenting on the quality of a specific core (or developer) here, just using an example).

You could play DoDonPachi for 100s of hours and not ever see some of the problematic areas with regard to slowdown. It gets really important in late second loop, and that likely takes some dedicated practice to consistently reach. Hopefully analysis of the various hardware gets you most of the way there, but you never know. You could likely hack of a 'debug' rom to jump you directly to late game, but some of these games have complex game states that may not be replicated by just 'jump to stage 2-5' so it wouldn't be a 'true' test.

The reality is that the pool of people capable of comparing 'full game' gameplay between FPGA and PCB runs are somewhat rare. It's especially bad now given the price involved with these games.

The best we can do is report differences/bugs as we find them and hope things get incrementally better over time.
suverman
Posts: 84
Joined: Mon May 25, 2020 1:17 am
Has thanked: 42 times
Been thanked: 6 times

Re: SHMUP Slowdown / Accuracy of Arcade Cores

Unread post by suverman »

PikWik wrote: Sat Jan 16, 2021 11:17 pm i do remember when CPS1 was being worked on, people were comparing the core to actual PCBs, but im not seeing this same dedication with arcade cores lately.
Its on going, maybe since nullobject doesnt have a central discord you miss it. There are some really good testers who are testing DDP core and have been testing it since beta 2. Dont worry about it.
User avatar
Chris23235
Top Contributor
Posts: 841
Joined: Sun May 24, 2020 8:45 pm
Has thanked: 106 times
Been thanked: 167 times

Re: SHMUP Slowdown / Accuracy of Arcade Cores

Unread post by Chris23235 »

nullobject explicitly stated in the Patreon post of the public release:
Known issues:
Slowdown accuracy. I haven't verified the accuracy of the slowdowns with a real PCB, but there are times when the game should be slowing down more (i.e. lots of sprites on the screen).
https://www.patreon.com/posts/dodonpachi-46150113

So this seems a non-discussion to me.
PikWik
Posts: 256
Joined: Sat May 30, 2020 7:00 pm
Has thanked: 192 times
Been thanked: 71 times

Re: SHMUP Slowdown / Accuracy of Arcade Cores

Unread post by PikWik »

zakk4223 wrote: Sun Jan 17, 2021 2:09 am The reality is that the pool of people capable of comparing 'full game' gameplay between FPGA and PCB runs are somewhat rare. It's especially bad now given the price involved with these games.
this was my thought as well and sparked some of my initial concerns.
having the PCB to tinker with and being a top tier player are 2 different things.
aberu wrote: Sun Jan 17, 2021 12:15 am 2. MAME, as a result of being an emulator in software that runs on PC's, has significant latency overhead, so the comparison is sorta apples to oranges. There may be some cores that will have temporarily some bugs, but the input lag of all software emulators should probably be counted into that comparison then. No emulators "feel" the same to me, and that's always been why it was hard for me to go through games entirely in emulators, and I've been using emulators since the mid-90's. Pretty much only modern-day versions of Mednafen with PS1 rpgs which don't require the tighter inputs have been the only thing that would feel okay.
i do want to mention groovymame again (especially with a CRT). that particular build of mame can output the original refresh rates of arcade games, and has some clever tricks to reduce input lag to imperceptible levels. its also one of the only ways to play the later cave CV1000 games at the moment; and of course the capcom CPS3 games, which wont have FPGA versions for quite a while.

for now, with the limitations of the MiSTer, everyone is stuck using emulators like groovymame for arcade emulation past the limits of the de10-nano. and to use a 'street fighter 2' to 'street fighter 3' analogy, playing the 1st gen cave games is very much like playing vanilla street fighter II, where once you experience all of the CV1000 shmups (or street fighter III 3rd strike), the upgrade is quite apparent.
the 1st gen cave games are obviously important to shmup history, but the later CV1000 cave games seem to be the gold standard for cave shmups.

and i understand these arcade cores are works-in-progress and will take time to get to a point of "perfection".
im here for the journey and i will absolutely be paying attention to what the developers are doing.
i have said this many times, but i believe FPGA is the future of retro gaming.

1s2aej7.jpg
1s2aej7.jpg (166.41 KiB) Viewed 9443 times
User avatar
aberu
Core Developer
Posts: 1144
Joined: Tue Jun 09, 2020 8:34 pm
Location: Longmont, CO
Has thanked: 244 times
Been thanked: 388 times
Contact:

Re: SHMUP Slowdown / Accuracy of Arcade Cores

Unread post by aberu »

PikWik wrote: Sun Jan 17, 2021 7:55 pm i do want to mention groovymame again (especially with a CRT). that particular build of mame can output the original refresh rates of arcade games, and has some clever tricks to reduce input lag to imperceptible levels.
Yes you are right that Groovymame can get imperceptibly close to the MiSTer in terms of input lag. IF you run Groovymame with a CRT, with run-ahead in many instances (which can often introduce major bugs), with having to set a specific frame delay which differs depending on game, crt, graphics card, display driver, etc... after you have bought one of a few specific ATI graphics cards that are no longer being produced for over 10 years (last I checked this is how it worked), and yes, then it is going to be within one frame just like the MiSTer.

However, for serial input that directly and as instantly as possible registers the input, and the video signals being produced are the same as original hardware, again directly from FPGA... That bit of lag is non-existent in the MiSTer in many cases. Lag tests in the NES core have shown that it's identical and in some rare cases actually faster than original hardware in response. Meaning sometimes less than a milisecond.

I'm not trying to say one is necessarily better than the other, but I think the MiSTer comes out on top in terms of value and ease to acquire the components and set it up. Also, if you aren't using a CRT and are using LCD, the MiSTer again comes out on top. Additionally, there just is lag that is going to be inherent with something like Groovymame due to it not having a serial connection that is directly interfacing with it's emulators, but it's imperceptible so it doesn't really matter :P
PikWik wrote: Sun Jan 17, 2021 7:55 pmfor now, with the limitations of the MiSTer, everyone is stuck using emulators like groovymame for arcade emulation past the limits of the de10-nano.
Right, but the thing is you have to remember, this is apples and oranges. The MiSTer platform and whatever platform it changes to in the future potentially (like if there is a new affordable dev board with a significantly more powerful FPGA on it), will always be limited in numerous ways to that of software emulators. You will probably not see a reimplementation of the PS2 in FPGA for... your lifetime, or if you do it will run on a 2000 dollar board 10 years from now or something :P.

Just gotta manage your expectations and recognize they are two different things.
birdybro~
akeley
Top Contributor
Posts: 1303
Joined: Mon May 25, 2020 7:54 pm
Has thanked: 416 times
Been thanked: 399 times

Re: SHMUP Slowdown / Accuracy of Arcade Cores

Unread post by akeley »

You can use Groovymame with just about any ATI card these days, modern ones included.

For sure, MiSTer is much better when it comes to single games, mostly in terms of simplicity because setting up a crtemudriver-capable rig is a lot of work. The zero-lag matters as well, though imo not as much as people like to paint it (the differences really are very low and many people can achieve great scores or complete 1CC runs on machines with some lag present - it's a question of getting used to it).

However, MiSTer's library is obviously just a fraction of the entire MAME set. Even if it keeps growing you will always need that second option, if you're a real arcade fan. For me, arcade in MiSTer is just a bonus, it's lovely to be able to play Defender or Neo Geo games but my core interest is in microcomputer and consoles, where MiSTer is a valid replacement for emulators.
rsn8887
Posts: 71
Joined: Sun May 24, 2020 9:48 pm
Has thanked: 23 times
Been thanked: 12 times

Re: SHMUP Slowdown / Accuracy of Arcade Cores

Unread post by rsn8887 »

It looks like input lag on Groovymame is quite bad, 2-3 frames compared to pcb, see this video for example:
https://youtu.be/t0fkVi6p1oE

I think MiSTer is already performing much better than that?

Also there have been tests of MiSTer compared to real hardware, for example with SF2 and they stayed perfectly in sync for a long time, at least 6 minutes or so. Has this been tested with Groovymame even?

https://youtu.be/sh3C_wrNnBg
dmckean
Posts: 307
Joined: Sat Jan 16, 2021 7:03 am
Has thanked: 387 times
Been thanked: 95 times

Re: SHMUP Slowdown / Accuracy of Arcade Cores

Unread post by dmckean »

rsn8887 wrote: Wed Jan 20, 2021 2:16 am It looks like input lag on Groovymame is quite bad, 2-3 frames compared to pcb, see this video for example:
https://youtu.be/t0fkVi6p1oE

I think MiSTer is already performing much better than that?
Yes, MiSTer is better than that. Lag can be as low as only an extra 1ms over usb (depending on implementation) and lower than that with SNAC or Bliss.

Also, since that video you linked to was made, MAME has added low latency mode that chops off one second of input lag across the board. So now it's 1-2 frames compared to pcb. That's really not too shabby and about as good as you can expect with the extra overhead of software emulation.
User avatar
aberu
Core Developer
Posts: 1144
Joined: Tue Jun 09, 2020 8:34 pm
Location: Longmont, CO
Has thanked: 244 times
Been thanked: 388 times
Contact:

Re: SHMUP Slowdown / Accuracy of Arcade Cores

Unread post by aberu »

rsn8887 wrote: Wed Jan 20, 2021 2:16 am It looks like input lag on Groovymame is quite bad, 2-3 frames compared to pcb, see this video for example:
https://youtu.be/t0fkVi6p1oE

I think MiSTer is already performing much better than that?

Also there have been tests of MiSTer compared to real hardware, for example with SF2 and they stayed perfectly in sync for a long time, at least 6 minutes or so. Has this been tested with Groovymame even?

https://youtu.be/sh3C_wrNnBg
On CRT, with a stick hooked up via something like SNAC or potentially LLAPI, and if the core is designed properly, the MiSTer can be basically identical to original hardware (under 1 ms input lag).

That being said, this video goes into zero detail as to how everything is hooked up. If the arcade pad is serial, that's going to make a good amount of difference. If it's usb but the polling rate is low, that's going to make a lot of difference. if they aren't using run ahead, that's going to make a lot of difference. Just to be fair to groovymame as a project, this youtube video doesn't really demonstrate anything since there's no information about their testing methodology.

I'm looking around for comparisons for groovymame and I don't see any yet which is weird. I'm sure someone's done them for timing before.
birdybro~
PikWik
Posts: 256
Joined: Sat May 30, 2020 7:00 pm
Has thanked: 192 times
Been thanked: 71 times

Re: SHMUP Slowdown / Accuracy of Arcade Cores

Unread post by PikWik »

http://forum.arcadecontrols.com/index.p ... c=160722.0
that has a proper input lag test with data results and a video to watch

VTlFJrf.png
VTlFJrf.png (133.94 KiB) Viewed 9125 times
PikWik
Posts: 256
Joined: Sat May 30, 2020 7:00 pm
Has thanked: 192 times
Been thanked: 71 times

Re: SHMUP Slowdown / Accuracy of Arcade Cores

Unread post by PikWik »

been following this dude for a little while after he started to test the dodonpachi beta cores

https://www.youtube.com/watch?v=lZEsi_Ncdjs

super good news.
the slowdown is accurate in nullobjects dodonpachi core !!

(im not 100% on the youtuber's comments about MiSTer supporting freesync, but still great news)
PikWik
Posts: 256
Joined: Sat May 30, 2020 7:00 pm
Has thanked: 192 times
Been thanked: 71 times

Re: SHMUP Slowdown / Accuracy of Arcade Cores

Unread post by PikWik »

oh wow, jotego just updated his CPS1/1.5 core with a .08% difference in DMA timing (slower) compared to a SF2 PCB.
this change will be most noticeable in the SHMUPs for CPS1, where the DMA timings work in combination to govern the "slowdown," but this update will also help other titles where small audio/visual garbage may occur, due to DMA timings in the CPS1 core being too slow (without jotego's turbo on) or too fast (with turbo on). DISCLAIMER - im only repeating what other people have said, and by no means a DMA timing master

https://www.patreon.com/posts/cps1-cps1-5-46899296

very cool to see this level of detail being put into an arcade core, and CPS1 definitely deserves the work.
i dont think i need to remind anyone of how important CPS1 is to arcade history, or how important SF2 is to fighting game fans.

and if you havent played this SF2 rom hack, you are missing out :!: - https://sf2mix.github.io/system.html
SF2mix is my new favorite way to play SF2. the extra moves, graphic changes, and slight adjustments to the mechanics create a modern re-imagining of SF2. this isnt a rushed janky rom hack, and some very careful and deliberate work went into making SF2mix.

if you run the update_all script, SF2mix will be in the SF2 champion edition alternatives folder
JonnyTenebrous
Posts: 18
Joined: Tue May 26, 2020 11:27 am
Has thanked: 22 times
Been thanked: 4 times

Re: SHMUP Slowdown / Accuracy of Arcade Cores

Unread post by JonnyTenebrous »

PikWik wrote: Mon Jan 25, 2021 10:49 pm (im not 100% on the youtuber's comments about MiSTer supporting freesync, but still great news)
I've got a freesync monitor, and the OSD tells me it's running at 57khz (OSD is rounding to the nearest integer, obviously) when I run Dodonpachi. And there is no tearing (which would definitely be visible without vsync enabled). So.... it appears to be doing its thing and running at the refresh rate of the input signal.
ash2fpga
Posts: 237
Joined: Tue May 26, 2020 6:20 pm
Has thanked: 62 times
Been thanked: 28 times

Re: SHMUP Slowdown / Accuracy of Arcade Cores

Unread post by ash2fpga »

JonnyTenebrous wrote: Sun Jan 31, 2021 9:14 pm I've got a freesync monitor, and the OSD tells me it's running at 57khz (OSD is rounding to the nearest integer, obviously) when I run Dodonpachi. And there is no tearing (which would definitely be visible without vsync enabled). So.... it appears to be doing its thing and running at the refresh rate of the input signal.
Which monitor are you using?
JonnyTenebrous
Posts: 18
Joined: Tue May 26, 2020 11:27 am
Has thanked: 22 times
Been thanked: 4 times

Re: SHMUP Slowdown / Accuracy of Arcade Cores

Unread post by JonnyTenebrous »

ash2fpga wrote: Mon Feb 01, 2021 1:01 am
JonnyTenebrous wrote: Sun Jan 31, 2021 9:14 pm I've got a freesync monitor, and the OSD tells me it's running at 57khz (OSD is rounding to the nearest integer, obviously) when I run Dodonpachi. And there is no tearing (which would definitely be visible without vsync enabled). So.... it appears to be doing its thing and running at the refresh rate of the input signal.
Which monitor are you using?
An LG 27UK650. 4k, IPS, 2018 model.
User avatar
lamarax
Top Contributor
Posts: 472
Joined: Wed Nov 11, 2020 6:28 pm
Has thanked: 33 times
Been thanked: 199 times

Re: SHMUP Slowdown / Accuracy of Arcade Cores

Unread post by lamarax »

Is freesync initialized differently than G-sync? Like, is it a "raw" capability of the panel? 🤔 Because for G-sync (which my monitor can do) to work, it has to be called by the GPU driver in the OS.
zakk4223
Posts: 270
Joined: Sun May 24, 2020 10:55 pm
Been thanked: 107 times

Re: SHMUP Slowdown / Accuracy of Arcade Cores

Unread post by zakk4223 »

The factory settings/debug menu of my LG will display both input sync rate and output sync rate. As long as you enable freesync on the monitor, they always match when using my Mister, even on cores that are non-60hz.

The gsync vs freesync question is complicated by the fact that 'gsync' means multiple things because nvidia likes confusing people.

However, in both cases the likely hand wavy answer is that the VRR part is a 'raw' capability of the panel (assuming you enable it in the monitor settings); the driver enablement is more about setting up the display driver section of the video card to internally handle VRR, negotiate min/max freesync ranges etc.
grogre
Posts: 3
Joined: Sun May 24, 2020 7:31 pm
Has thanked: 1 time
Been thanked: 1 time

Re: SHMUP Slowdown / Accuracy of Arcade Cores

Unread post by grogre »

PikWik wrote: Sun Jan 31, 2021 7:39 pm oh wow, jotego just updated his CPS1/1.5 core with a .08% difference in DMA timing (slower) compared to a SF2 PCB.
this change will be most noticeable in the SHMUPs for CPS1, where the DMA timings work in combination to govern the "slowdown," but this update will also help other titles where small audio/visual garbage may occur, due to DMA timings in the CPS1 core being too slow (without jotego's turbo on) or too fast (with turbo on). DISCLAIMER - im only repeating what other people have said, and by no means a DMA timing master

https://www.patreon.com/posts/cps1-cps1-5-46899296

very cool to see this level of detail being put into an arcade core, and CPS1 definitely deserves the work.
i dont think i need to remind anyone of how important CPS1 is to arcade history, or how important SF2 is to fighting game fans.

and if you havent played this SF2 rom hack, you are missing out :!: - https://sf2mix.github.io/system.html
SF2mix is my new favorite way to play SF2. the extra moves, graphic changes, and slight adjustments to the mechanics create a modern re-imagining of SF2. this isnt a rushed janky rom hack, and some very careful and deliberate work went into making SF2mix.

if you run the update_all script, SF2mix will be in the SF2 champion edition alternatives folder
I have a question about turbo mode in cps1 core, it is told that in cps1 boards the olders run the motorola 68000 at 10 mhz, later, since street fighter 2 the new boards run the 68000 at 12 mhz. So for the cps1 core on mister turbo mode is 68000 at 12 mhz or is it something else ? In that case the way to play SF2 and later games on CPS1 should be with turbo mode on and not in normal mode, which are the correct settings in this case, should we consider the turbo mode=68000 processor at 12 mhz ?
PikWik
Posts: 256
Joined: Sat May 30, 2020 7:00 pm
Has thanked: 192 times
Been thanked: 71 times

Re: SHMUP Slowdown / Accuracy of Arcade Cores

Unread post by PikWik »

hey there, thats a great question, and one that i was wondering about as well.
you are correct, starting with SF2 champion edition, CPS1 PCBs were installed with 12mhz CPUs to support the later titles, versus the original 10mhz CPU used for the original SF2.

https://www.arcade-projects.com/threads ... have.13927

https://www.twingalaxies.com/yesaffinit ... z-vs-12mhz

there are some great discussions there if anyone wants to dive deeper into the subject.
from reading those, 12mhz has better input detection but may be more difficult due to the AI reading moves quicker,
but playing SF2 CE and HF in 10mhz will be a tad slower and allow some combos that arent possible on a 12mhz CPU.

even twin galaxies hasnt landed on a consensus, and world records are submitted with both types of boards for the "same" game.

TL:DR version is to just leave jotego's turbo mode on, as 12mhz is how most people with PCBs play those games.
(but im still not 100% on what the turbo mode is doing)
PikWik
Posts: 256
Joined: Sat May 30, 2020 7:00 pm
Has thanked: 192 times
Been thanked: 71 times

Re: SHMUP Slowdown / Accuracy of Arcade Cores

Unread post by PikWik »

and regarding the freesync/gsync support for MiSTer.

i think having a TV/monitor with freesync is definitely a good idea, and in a few years time i imagine most modern TVs/monitors will have this technology. it certainly wouldnt hurt to seek out a TV/monitor that supports the widest range of refresh rates, yes
XtraSmiley
Posts: 109
Joined: Wed May 27, 2020 11:33 pm
Has thanked: 101 times
Been thanked: 15 times

Re: SHMUP Slowdown / Accuracy of Arcade Cores

Unread post by XtraSmiley »

PikWik wrote: Tue Feb 02, 2021 12:05 am
TL:DR version is to just leave jotego's turbo mode on, as 12mhz is how most people with PCBs play those games.
So two things, 1. is there any issue with earlier games not designed at 12mhz having problems of some kind? And 2. He should rename them to 10mhz mode and 12mhz mode instead of turbo on/off.
PikWik
Posts: 256
Joined: Sat May 30, 2020 7:00 pm
Has thanked: 192 times
Been thanked: 71 times

Re: SHMUP Slowdown / Accuracy of Arcade Cores

Unread post by PikWik »

that is probably a better question for someone who has those PCBs and more experience with CPS1 (not me)
but, i did find this post by someone on an arcade forum.
Upgrading a 10MHz PCB to 12MHz
According to Apocalypse even if you do have a 10MHz board it can be upgraded to a 12MHz by replacing the crystal.
Most games should work on either board, but there will just be more slow-down on the 10MHz boards.

Here are the different games and which board they're compatible with:

10MHz boards
1941: Counter Attack
Captain Commando
Carrier Air Wing/U.S. Navy
Dynasty Wars/Tenchi wo Kurau (12MHz version also available)
Final Fight
Forgotten Worlds/Lost World
Ganbare! Marine Kun
Ghouls'n Ghosts/Daimakaimura (12MHz version also available)
The King of Dragons
Knights of the Round
Magic Sword: Heroic Fantasy
Mega Twins/Chiki Chiki Boys
Mercs/Senjou no Ookami II
Nemo
Pokonyan! Balloon
Street Fighter II: The World Warrior
Strider/Strider Hiryu (12MHz version also available)
Three Wonders
U.N. Squadron/Area 88 (12MHz version also available)
Willow

12MHz boards
Adventure Quiz Capcom World 2
Area 88 (10MHz version also available)
Daimakaimura (10MHz version also available)
Mega Man: The Power Battle/Rockman: The Power Battle
Pang! 3
Pnickies
Quiz & Dragons: Capcom Quiz Game
Quiz Tonosama no Yabou 2: Zenkoku-ban
Street Fighter II': Champion Edition
Street Fighter II': Hyper Fighting
Street Fighter Zero
Strider Hiryu (10MHz version also available)
Tenchi wo Kurau (10MHz version also available)
Varth: Operation Thunderstorm

CPS 1.5 boards
Cadillacs and Dinosaurs
Muscle Bomber Duo
The Punisher
Saturday Night Slam Masters/Muscle Bomber: The Body Explosion
Warriors of Fate/Tenchi wo Kurau II: Sekiheki no Tatakai/Sangokushi II
PikWik
Posts: 256
Joined: Sat May 30, 2020 7:00 pm
Has thanked: 192 times
Been thanked: 71 times

Re: SHMUP Slowdown / Accuracy of Arcade Cores

Unread post by PikWik »

and yes, i do think "turbo" is confusing.

im not exactly sure what turbo mode is enabling tho
ExCyber
Posts: 217
Joined: Sun May 24, 2020 3:33 pm
Has thanked: 11 times
Been thanked: 66 times

Re: SHMUP Slowdown / Accuracy of Arcade Cores

Unread post by ExCyber »

Turbo doesn't change the clock speed, it bypasses CPU bus arbitration for sprite DMA. Both of these approaches have the effect of running more CPU instructions per second, but the underlying mechanisms are pretty different.
PikWik
Posts: 256
Joined: Sat May 30, 2020 7:00 pm
Has thanked: 192 times
Been thanked: 71 times

Re: SHMUP Slowdown / Accuracy of Arcade Cores

Unread post by PikWik »

ah hah!

k9kAI40.png
k9kAI40.png (44.19 KiB) Viewed 5608 times
KN6c3qc.png
KN6c3qc.png (39.59 KiB) Viewed 5608 times

also, i found out the japan resale mras (in the alternatives folders on your rom folder, if you used the update_all script) are 12mhz versions of the CPS1 titles.

these include - Dynasty Wars, Area 88, Tenchi wo Kurau, Varth, Daimakaimura, & Strider
Post Reply