Using the HIgan core gives me pretty bad audio stuttering ingame. I didn’t experience this even with the accuracy bsnes core. I tried a few settings like lowering the internal resolution and switching the video driver but it doesn’t help. This is with and without shaders.
EDIT: I managed to fix it. Updated RA and the stuttering seems to be gone now. Great job on the core by the way!
Nice! All special-chip BIOS, included the SGB, is working. Only 3 things:
The over-clock bug of the original Super GameBoy 1 is present in the SGB1 and SGB2 modes. In the real SuperGameBoy 2, this is fixed.
The soft-reset is gone. The reset option in the Quick Menu only do a hard-reset, like if I turned off the SNES and turned on again.
Is there anyway to auto-detect the resolution like the bsnes/bsnes-mercury 0.94 cores? The shaders are looking ugly with manual 512x448 internal resolution in the games that are 256x224.
I think this are present in Higan v105 itself but I report this issues here anyway.
Thanks!
I really like the idea of renaming bnes-mercury to just “mercury”! It would make the Online Updater less overwhelming. When I first started using RetroArch for Windows ( I was only aware of the 3DS version back then lol ) I didn’t know what SNES core I should use!
Is it possible that the core is slower that standalone Higan?
I get 40fps in Yoshi’s Island in the spinning island screen, while in Higan I get 60fps (sync audio).
When I fast forward in this screen I have 67fps in both Core and Standalone, so it seems the Core has some speed issues.
i7 3770k at 3.9GHz stable no downclocking (no C states).
I had something like that before, and I either increase audio latency or use threaded video on my low system laptop. Not sure about the case of this core.
Did you try disabling vsync in RetroArch and fall back to audio sync? Seems like that would be a more apples-to-apples comparison.
I can’t speak for the standalone version of higan but now that you mention it I could almost swear that the slowdowns I experienced during the boss fight in Akumajou Dracula - Orphic Vipers (Twin Head Dragon) in Stage 3 earlier today isn’t something I have experienced before.
It showed 60fps but parts of the fight felt like snails pace which I thought was really odd and I got raped completely because of this.
I’m on a i7 4790K at stock clock but the Turbo Boost goes up to 4.4GHz and that should be well enough for higan, I don’t think I’ve had problems with speed on bsnes Accuracy v094 but I could be wrong.
I probably have to make more tests but damn they’re so time consuming, there should be more hours per day…
Yep, that did it, Thanks huterk!
Although vsync is important for that smooth experience which I consider crucial part of the quality of emulation.
I want to actually forget I’m playing an emulator and concentrate on the game, so video stuttering or audio noises are a big distraction for me.
I’d rather play a lower accuracy emulator but have a smooth experience… like the hardware console.
Yeah, I wasn’t actually suggesting keeping it that way, just trying to suss out whether the libretro core was behaving differently from higan standalone. I’m glad to hear they’re similar when using the same syncing method.
@SigmaVirus v105 is significantly slower than v094, yes. I forget all of the performance-affecting things that byuu has changed since v094, but I recall one single change had something like a 10% hit (might have been the bignum/femtosecond math…?).
Once hex_usr gets nSide ready, we’ll get a v105 balanced core which should be very popular.
But there are not a real, i mean, visible or playable difference in these cores afaik in Snes emulation (i am talking about v094 accuracy). Or i am wrong?
That’s correct, though you could say the same thing about v094 vs snes9x in most cases.
One noticeable thing this new core does better than Snes9x or the old BSNES cores is transparency blending when outputting 256x224. If you set the internal res to that and turn on the blur core option, the foreground transparency effects in Kirby’s Dreamland 3 (check level 1-4) look like real transparency instead of weird “every other line” fake transparency. And it works well with shaders at that internal res.
Yes, is noticeable in Jurassic Park (HUD and other things). But my machine cant run higan 105 core. Are other alternatives to see this effect?
There are a lot of shaders that will do it. snes-hires is the most direct.
Wow, i didnt know that. Now works perfect.
Ahh… choices choices.
The clear and conclusive answer for that old debate would be as following:
Once CPUs are powerful enough to effortlessly and run cycle accurate emulation (ie. not with 95% cpu usage like Higan), then cycle accurate emulation will be the first option to go for.
But, since Higan barely reaches 60fps in some cases on a modern i7, it is a secondary option for me reserved for testing.
Yep, that’s a prudent plan, I think, and it’s what I typically follow myself. Super Mario World runs essentially “perfectly” in snes9x, so it’s just burning cycles to run it in higan-accuracy. That’s not always the case for more obscure and/or complex titles, though, and being able to completely rule out emulator quirks can be valuable for debugging purposes.
keep the accuracy core, if there are still possible optimizations that can be done, else well hear cries about not having an “accurate” core when there is one available…
(the core kinda stutters for me too, but i only tested a few)
The over-clock bug of the original Super GameBoy 1 is present in the SGB1 and SGB2 modes. In the real SuperGameBoy 2, this is fixed.
According to here: https://higan.readthedocs.io/en/latest/guides/import/#super-game-boy-games, indeed it has not emulated timing properly yet:
higan supports the Super Game Boy 2 base cartridge, so you can use the extra borders, but does not yet emulate the timing change so games still play slightly too fast.
Is there anyway to auto-detect the resolution like the bsnes/bsnes-mercury 0.94 cores? The shaders are looking ugly with manual 512x448 internal resolution in the games that are 256x224.
I also strongly and desperately need this. Looking 240 games at 480 is indeed makes shaders ugly especially for shaders emulate scanlines (like CRT-Royale). It makes scanline too thin to see (or not seen at all): 240 at 240, 240 at 480.
In addition, looking 480 games at 240 is extremely distorted. Please be careful for NSFW little bit (Sorry but this one has the most obvious distortion) 480 at 480, 480 at 240. Look how distorted the name is.
Switching internal resolution manually is so painful. I really desire auto internal resolution just like v094 libretro version.