[Pi 4] Cave CV1000 Performance Thread
-
I'm not sure if this option gets much popular use, but I found out you can just create sub-folders inside a system's rom folder and they show up as an entry in the rom list for that system.
So if you make a folder under arcade or fba called 'cave' you can put all your cave games in it. Then all cave games will be inside an entry called 'cave' in the rom list. Then you can add symbols in front of the name if you want to make it appear at the top.
I am considering trying this as a way to categorize all the arcade roms on my setup.
-
@YFZdude Sure, you can do this if you want to physically move the ROMs. Scraping may not necessarily follow this "unknown" directory structure without manual edits to files. It depends on your situation I guess.
-
Think I'll just follow YFZdude's suggestion and simply use a subfolder.
I do that with most of the systems anyway since I don't scrape any of the contents in them.
My folders on each are usually something like:
Licensed
Translated
Unlicensed (with even more subfolders in there) - for romhacksIt's nice to have a tree structure. :)
-
Dink updated fbneo a few days ago with more speed hacks for CV1000. Testing SDOJ doesn't reveal much of a difference for me so far, but it's always appreciated nonetheless! Figured you gents would like the update.
-
@dodonpachi said in [Pi 4] Cave CV1000 Performance Thread:
Dink updated fbneo a few days ago with more speed hacks for CV1000. Testing SDOJ doesn't reveal much of a difference for me so far, but it's always appreciated nonetheless! Figured you gents would like the update.
For SDOJ (and the Knuckles hack) the sweet spot for me on an overclocked Pi4B is to set the CPU clock to around 50%. The red character's bomb is otherwise always going to be a problem with framerate. Was able to use a different character and play through the entire first level without any significant framerate hit.
I think it's a small miracle these Cave games are running so well -- a testament to the hard work of the fbneo crew.
-
While I previously mentioned that I didn't notice a difference with the speed hacks, it must have been because the binary wasn't updated to match that version or something. I just updated lr-fbneo now, and played through SDOJ with incredible performance. Using el rikas latest suggested CPU and blitter settings (43.2 and 57), I encountered only two instances of sound drops, at the heaviest part of the final stage. Even then, it was only for a split second.
I'm floored at what a difference lr-fbneo has made so far. When I first made this thread back in April, I was lucky if I heard the entire stage 1 theme without drops, using lr-mame2016. Now I just played through the whole game at nearly full speed.
-
Hi all. Thanks to everyone who got this working so well, and it's been wonderful being able to play all these cave shooters on a pi. I know I should probably join the FBN forum and ask there, but as I'm new to all of this and have an account here I thought I'd ask: has anyone figured out good numbers to get Pink Sweets running well? On the linked thread that was still discussed as a work in progress, but I haven't seen any follow-up. I may end up testing around myself if nobody else has.
Cheers!
-
@bbilford83 Try 37% CPU, with no blitter delay. That's the recommended value in MAME (on a good PC)
-
@dodonpachi Thank you! Just played the first level and it seemed perfect!
-
@bbilford83 glad I could help!
-
@dodonpachi I might as well also ask, I've been messing around with getting the FM Towns Marty emulated on my pi 4 (also a system with a lot of good shmups) and it seems most everything runs like garbage without underclocking the CPU, but doing so has amazingly positive results. But I've been tinkering game-by-game because some games run best with really heavy CPU underclock (like under 40) and some with nearly none. I'm not familiar with whatever list of mame recommendations you found that gave the CPU underclock for Pink Sweets - is there something like that for non-arcade systems I could check to see recommendations for FM Towns? I assume not, but if so, it would be great not to reinvent the wheel...
-
@bbilford83 unfortunately, I wouldn't know. The Cave CV1000 games are a unique situation. When Cave developed that series of games, the hardware they used was strategically underpowered so that the game would slow down when there were too many bullets on screen. MAME dev was able to emulate the hardware almost perfectly--the only thing they haven't figured out yet is how to accurately replicate the throttling of the CPU during certain segments of the game. Because Cave's games have a niche following among enthusiasts, they came up with these CPU/blitter values as a sort of workaround for the missing emulation. Not only does it somewhat replicate the slowdown, it has the incidental benefit of making the games more playable on lower end hardware.
-
@dodonpachi That makes sense. It did seem implausible to me that all mame/fba would have similar info out there but I didn't know the backstory. In any case I did some testing and just posted my results in the thread about fmtmarty. I doubt they're as accurate as these but they got all my games solidly playable.
-
Sorry to jump in here. I’m running a Pi4 with overclock set to 1800mhz. I’m using FBneo all the CPU clock and blitter settings as recommended for the emulator here:
https://neo-source.com/index.php?topic=3934.0
I’m not sure how to tweak the GPU as noted by some members in this thread.
The problem is, while the slowdown seems more accurate with the CPU clocking and el_rika’s tenth percent/blitter values, it seems like I’m getting a lot more audio stuttering and big frame drops than people in this thread, and I’m not sure what I’m missing. Someone here noted they only got one instance of audio stuttering in STG4 of Daifukkatsu, but I’ll get instances in the first stage during big explosions where the audio stutters and slows down, and when every end of stage boss blows up and releases medals. It’s really prominent in Akai Katana, stuttering even on the character select screen and during stage 1, from the mid-boss onward.
Can someone give me a hint to try resolving this so I can match the results you guys have achieved here. Should I be using GPU/CPU sync in Retroarch menu? Should I be overclocking the Pi more than 1800mhz?
I’ve tried tinkering with audio latency too, but I find if it goes up too high it actually makes the game itself choppier and everything worse.
Thanks for your help!
-
@skykid My pi4 is at cpu 2000, gpu 750. I've seen those numbers floating around for years so I kept using them.
I'm not sure if it's a placebo effect or not, but I believe I get slightly better speeds using bullseye/64bit but there's no official way to do that right now.
Also, forget about Akai Katana. It's too slow in certain sections even with the settings I'm using. Maybe the people doing extreme overclocking past 2100 can get better speeds but I'm not doing that.
-
Ah interesting. I didn’t see any other notes on Akai Katana being slow. For me it’s not too slow to play, and responds well for the most part, but I get big drops in certain sections.
Overclock is listed as “extreme” at 1950mhz and “insane” at 2100. I wasn’t sure if I should risk that. I’m using an Argon One case which acts as an entire thermal heat sink enclosure, but I wasn’t sure about pushing it.
EDIT:
Ok I overclocked to 1950mhz and that seems to have brought it up to speed (excuse the pun). Performance is vastly improved and on testing I can’t hear anywhere near the same audio stuttering as before. I didn’t set the GPU but I have a feeling that when the mhz are overclocked in the front end it sets the GPU automatically.
I’ll keep testing but it seems great so far!
-
You want max FPS in every situation - try this -> :D
-
@skykid I also have a Argon One case and I have my Pi4 overclocked to 2100 MHz without issues.
There is no real risk in overclocking the CPU/GPU as the Pi automatically throttles the frequency when it overheats to more than 80°C.
There are different load tests out there, where all four CPUs of the Pi4 are working at full load. With that you can see if the system stays stable at a temperature lower then 80°C.... but even then this should be no problem, as in emulation there are seldom all four CPUs used at full capacity at the same time ...
-
@sirhenrythe5th Is that yours? :D
-
@ecto no, even if i like the idea.
Similar to a calculator with SSD-Harddisk :D
Contributions to the project are always appreciated, so if you would like to support us with a donation you can do so here.
Hosting provided by Mythic-Beasts. See the Hosting Information page for more information.