Beetle Saturn

Hi, after few months i tried again to play some saturn, I tried mednafen stand-alone and the retroarch core and for me there is a big difference in performance, in the stand-alone i have constant 60 fps with some dips to 58 in some games (maybe due to the saturn itself) but with the beetle core it struggles to reach 50…

a tried few games 2d & 3d, same results: 2 examples

Ninpen Manmaru (Japan) -> 3d game, stand-alone 60fps rock solid, beetle struggle in late 40’ fps

X-Men vs. Street Fighter (Japan) -> 2d game, stand-alone 60 for the most part some dips 57-58 (music and sound no problems), in Beetle 45 fps with audio slowdowns and crackles.

I tried all the video drivers (vulkan, gl, glcore, d3d11) and all audio drivers: no luck.

Am I doing something wrong? is there a WIP newer version of the core (the libretro core I have from online updater is 1.22.2 (2 years old?) the last mednafen is 1.27.1 )?

I’m running windows 10, i5 6400 and GTX 960.

I can run GT4 with pcsx2 libretro’s core @1080 60fps rock solid with d3d11

It’s been a few years since this core wasn’t updated to keep up with Mednafen’s. I believe you want to use it in Retroarch to benefit from the frontend’s options, shaders, etc, but, like you said, there’s a big performance hit. Have you tried downloading the latest RA and running Bettle Saturn from there using everything in its defaults? If so and you still notice lower performance, it’s probably the way it will run no matter what you do for now. There are other threads here where people often compare standalone emulators with their core counterparts, like PPSSPP, Dolphin and now PCSX2, usually there are comparisons with framerate and input lag added on Libretro’s side. I believe it’s difficult to maintain all cores and keep up with their standalones. But I also believe there’s always ways to improve performance and sometimes even surpass the standalones, like it happened to other cores. Still, since there’s huge drops in your tests, please let us know if you try the defaults in a separate, like, the full latest Retroarch extracted to another folder, so you can be sure there’s nothing in your current settings conflicting somehow.

1 Like

hi, thanks for your answer,

To answer your question: yes! I forgot to mention it in the post, but after a few tries with “my” retroarch folder I tried a fresh install, all default, but no differences in performance, in the freshly extracted retroarch I also did all the drivers tests (vulkan, gl, glcore and d3d) taht i mentioned above.

I know the difficulties to port and keep up with the upstream, mine is not a complain, but recently I did the switch from pcsx2 stand-alone to the libretro port and I was very happy with the performance, also duckstation is going great and I stopped using the stand-alone as well…so I tried also the beetle saturn core hoping in some good news… Maybe ps1 & 2 are much more popular than the saturn :((

well, I guess at least I know that my install of retroarch does not have problems. ^^

I’ll test the same games you mentioned above later and I’ll come back with my results, the CPU I’ll be testing it with is a i7-2600 4ghz+

It will take some time, though.

1 Like


So, in ASrock A-tuning Utility I changed from standard to performance and the framerate went up, with this configuration it behave just like the stand-alone: framerate went up and it stays on 60fps (with 1-2 fps drops depending on situations)… I will say that I usually keep my CPU in “standard mode” and framerate is good with pcsx2 core, and Mednafen stand-alone. From what I saw, CPU usage, on “performance” mode, does not go over 65% while playing with Beetle Saturn.

There maybe be something related in how beetle Saturn “draws” resources form the CPU?

1 Like

That’s funny because recently I ran Beetle Saturn on my Android phone and video seems to be smoother than what I remember since last time (5-6 months ago). There is still some crackling audio, but the FPS is much closer to playable.

Looking at the Github repo it seems there were some updates in the last months, but this doesn’t look performance related.

So, am I imagining things or the core’s performance improved recently? :thinking:

PS. I’m using gl driver and I don’t think I touched any video settings since then (not many opportunities to play recently).

So, I uninstalled the asrock utility just to always have the CPU at 3.2 GHz,

X-Men vs. Street Fighter (Japan) seems to keep almost constant 60 fps, so does Ninpen Manmaru.

Now I just tried Marvel Super Heroes vs. Street Fighter already frame drops in the title screen, and in the character selection screen, then then when the round start the frame rate is unstable ranging from 52 to 60 FPS, tried in the stand-alone and it doesn’t move from 60 at all times.

can somebody try Marvel Super Heroes vs. Street Fighter (jap)?

Here are 4 tests in 2 games:

Mednafen Saturn running Pandemonium!

Beetle Saturn running Pandemonium!

Mednafen Saturn running Burning Rangers

Beetle Saturn running Burning Rangers

Results are as expected, although, I think in both cases Pandemonium doesn’t sound quite right, Mednafen runs both games without problems, specially Burning Rangers, frame is stable and frame pacing is really noticeably unstable in Beetle’s.

As a bonus, MedGui Reborn has a “Test your CPU” option and it showed my 10yo CPU is still a beast.


thank you to take the time to do those tests!

Well, so I would say that the RA’s core is good but still not there yet… you have a slightly higher freq CPU and frames are dropping as well, so there is probably nothing wrong with my config/set up, the stand-alone is still the best choice at the moment for me, sadly :((

Just out of curiosity, if you have the chance, try Marvel Super Heroes vs. Street Fighter (jap)

x-man vs SF have a decent framerate but MSH vs SF, drops many frames, and they are almost the same game…

in mednafen the run both at constant 60, in beetle have different behavior, very odd.

you said it was fine when your CPU was in “performance” mode? It’s a very common thing for strong CPUs to finish the emulation task with enough time to spare that the OS’ power-savings routines see the idle time and then down-clock the CPU, making it too slow to run full-speed until it up-clocks and the process begins again.

I just tried, here are the results:

Mednafen running Marvel vs Street Fighter

Beetle Saturn running Marvel vs Street Fighter

Mednafen often keeps 60FPS but there are some weird sound skipping at times, while Bettles won’t even reach 60, there’s something weird about this game being emulated, what it is, I don’t know. In any case, one day, when Mednafen Saturn becomes stable and developed enough to have a dynarec, then we can say it’s a good enough option for action/fighting games, because, even for those that actually reach and is 60 at a stable level, like you mentioned X-men vs SF, you still have an underwhelming amount of input lag, which obviously ruins any games like these and fortunately you can always play the original arcade version in MAME and FBNEO with runahead, eliminating input lag, games running ALWAYS at their original intended speed/pacing. Don’t think that my CPU having slightly higher clock even make a difference, these emulators aren’t mature enough, even modern CPUs won’t run these as smooth as you think, it’s a software level development issue to be solved. The PS1 is a well known system to emulate, but even if you choose the software variation and crank up the internal resolution from 4x or higher, no CPU will be strong enough to keep up. Don’t tell me you really want to play Norimaro at 60 FPS on Saturn, c’mon, he’s not that big of a deal, don’t hold your breath, the real hardware is better than Mednafen. Though it would be amazing if Shima really had released SSF’s source code, but it seems he held it back.

Performance seems pretty similar with both here (xmen vs sf; haswell laptop, linux). I don’t know how to show FPS in mednafen, though, so it could be off by a few FPS. Hiddenasbestos’ rebase branch seemed a little smoother than either standalone mednafen or the buildbot core here, but it was still hovering around 56 FPS.

Some things to watch out for when making comparisons: make sure rewind, runahead and hard GPU sync are not enabled in RetroArch, as all of those can have a significant effect on performance. Likewise, make sure you don’t have frameskip enabled in mednafen.

1 Like

Yes, I press the menu button often in the videos to show no latency or other performance related settings are active, I actually never even enabled rewind, never used it. I also used all other settings in its defaults, mednafen is set to never skip frames. MSHvsSF didn’t run perfectly like it shows in the video on mednafen either, but since libretro’s core is so far behind, it shows in all tests I made, it was mostly at 60 in the standalone mednafen fps but I played this game to death on the real hardware and I notice, specially input lag, the game just runs like a dream on the machine.

1 Like

@hunterk yes xmen vs sf runs at 60 also in beetle, the weird one is marvel super hero vs sf, as shown by aorin1 drops frames a lot and is almost the same game…

@aorin1 thanks for checking it man, yes if I want to play those fighters I will do it on fbneo, but it was a way to see where we are at with saturn, and as you said emulation for this machine is not there yet, but the stand-alone I think is good enough for slow-paced games (–> not for shooters and fighters).

1 Like

I was about to test the same games in SSF, no surprises here, still the best Saturn emulator around, there’s a reddit thread where a guy talks about using RTSS scanline option to reduce input lag further.

Marvel vs Street Fighter

Burning Rangers



I remember looking into compatibility lists last year and seeing mednafen had now the highest compatibility (idk about the beetle fork though).


I believe Mednafen’s focus on accuracy will make it the best option in the future, when it integrates dynarec. I made tests in previous posts playing 3 games, as far as playability goes, it’s SSF>Mednafen>Beetle respectively. I still think SSF compatibility and accuracy is very high, if you compare MSHvsSF across the 3 emulators, the speed and fluidity of the gameplay is really smooth on SSF and it never dropped a single frame while running any of the 3, even input lag is reduced there.

The bottlenecks for saturn emulation are the multiple vdps, furthermore sega saturn is really tight about cpu cycles, changing anything about the interpreter might require a major rewrite of everything related to timings.

I think SSF is using hardware-acceleration, but considering how hard it is to properly emulate the vdp1 with modern gpus, i’ve been wondering if it’s being emulated properly. A very good test for that is sega rally’s attract mode : proper vdp1 emulation will have no holes on the road and barely-visible dots on the borders, like this :

Not like this :


That’s true, SSF shows how my GPU usage increases when I disable v-sync and it goes wild, almost quadrupling the original hardware framerate, and it sure doesn’t look 100% right in the second screenshot you posted.

I was about to test the same games above in Kronos, I also tested Burning Rangers on Yabasanshiro so I can finally call it a day, testing these games across emulators that are being currently developed.

MSH vs SF running on Kronos

Pandemonium! Running on Kronos

Burning Rangers running on Kronos

Burning Rangers running on Yabasanshiro

I was surprised how differently Yaba runs this game against Kronos, it also seems to have less input lag, there’s something wild going on when Kronos try rendering this game, there’s a shadow of the main character rendered in walls, at times, I noticed when it happens the game almost crashes and crawls. Sometimes I enable fastforward and it runs above 70 FPS depending on the scene. In Yaba, the game is not 100 rendered properly, it’s darker than it should be, in both cores, some sound effects aren’t played correctly on both as well, like in Shining Force III games, most of them sound off. The door closing sound is off, SFIII magic sounds, etc. It’s amazing how this console’s architecture defies emulation as it defied developers back then.

Checking back the videos, YT is ruining the originals, I’m posting the latest ones in Gdrive, in case someone would like checking the original videos:

Yeah, Kronos keeps getting regressions like those since it started dealing with timings (which is why i was warning about changing interpreter above), hopefully it’ll get better with time.