Yes.
I’m not sure what do you mean by that? on my system that is clearly not the case:
Regarding input lag, as I said I cannot test it properly but Ymir does feel more responsive.
did you test it with a camera or by frame advance?
1 Like
If it runs on a dual core at normal speed, on your i5, it has plenty of resources to spare. I am currently testing it on an i7 2nd generation and it is working very well, although I have other configurations available.
I read that you are modifying the power options; RA has game mode enabled.
Could that have an impact?
Advancing the frames.
I can run (most games) with beetle but what i’m saying is that Ymir have better performance:
If you can, try Marvel Super Heroes vs. Street Fighter, on my system (in performance mode) can’t quite reach full speed.
If it does, it’s not noticeable, I’ve enabled this option and couldn’t tell any difference, nor even on the game’s performance.
Unless we’re talking about a scientific test, both original consoles with a CRT TV, we’re just comparing apples to oranges, comparing emulator to emulator isn’t accurate.
Street Fighter Alpha 3 Max and Metal Slug XX, both for the PSP has noticeable input lag on original hardware, this is mostly because neither are original PSP titles, Metal Slug was made for the NDS as Metal Slug 7 and ported over, if you play the NDS version, you’ll notice it it just plays much better. In the case of SFA3, Capcom is known for using some sort of translation layer, the game just isn’t running natively at all.
Actually, many Capcom arcade games have intended input lag to give an extra “weight” to the character animations. You know that slowdown that occurs when a character is hit by a fireball in Street Fighter 2 and its updates while in the air? It’s intended for a dramatic impact, it’s a choice.
A nice example of inherent internal input lag is loading Quake II for the PS1, any emulator which allows for overclocking the virtual PSX CPU will make the game run smoothly at 60FPS, as it’s unlocked but rarely reaches a steady 30 in the console. With that, you’ll also notice how leagues better the input feels during gameplay, this lag is caused by the game’s engine and the lack of performance due to the hardware limitations.
I’m assuming you tested it on real hardware, is that correct? It not, are we taking conclusions from imperfect emulators and lag inducing overhead? My brothers and I owned both the Saturn and PS1 in the late 90’s for years and input lag was never an issue, played tons of Pocket Fighter and SFZ3, along side many SNK and other Capcom titles.
OK, so, you tried on emulation, that can’t be taken as a solid opinion.
I’d love seeing a video footage of that, doesn’t sound correct, but if you record using a phone, showing you can run Mednafen Saturn on a Pentium 4 at full speed, you’ll impress me. SSF is the fastest Saturn emulator by far, I’m not even considering Yabause as it’s way too inaccurate in comparison.
This is also incorrect, I know the Saturn and I’m used to Saturn emulation for over a decade and I can assure you that Ymir feels way more responsive than any other Saturn emulator out there.
2 Likes
Do you still have access to the original hardware? If so, you could try a decent test using the 240p test suite. They have a lag test consisting of a stopwatch which measures frames. If you record that on slow-motion, we could have a more precise picture here.
It hardly is. For years, I’ve seen people often blaming their failures on negligible lag. There are games I never beat during my childhood, whereas now I can beat using the supposedly bad emulation. In my case, I don’t trust myself to remember frame accuracy, just a general feeling of responsiveness which may or may not conform to reality; sometimes, I just need to get used again to a game. As usual, your mileage may vary.
1 Like
I’ve already explained that the goal of the frame advance test is only to test the inherent lag of the game, not to get an accurate measurement of any real world lag added by the emulator and/or system.
OK, so, you tried on emulation, that can’t be taken as a solid opinion.
And yet you wrote this:
My brothers and I owned both the Saturn and PS1 in the late 90’s for years and input lag was never an issue, played tons of Pocket Fighter and SFZ3, along side many SNK and other Capcom titles.
I’ve said how to properly test for lag. High speed camera on console then do the same for the emulator on PC. Your memories from the 90s are irrelevant here and certainly not a solid basis on which to form an opinion or a good test of lag on hardware vs emulation.
Frame advance is just a quick and easy way to rule out large inherent/internal lag in the game as a cause of noticeable lag. Again, it will not measure real world, external lag caused by the emulator or the PC hardware, monitor, video options etc.
1 Like
Unfortunately, I do not own original hardware any longer, if I did, I’d meticulously test a few games and compare them to both Mednafen and Ymir, so, we’d have more down to earth results.
Same here, it’s not that I really remember, but it never got in the way, for instance, when I loaded up SFA3 on Ymir, I finished World Tour right after loading the game, just to test it, where on other emulators, including Mednafen and SSF, it just never felt good enough for me to play. A non-scientific test for me is trying Guy’s Final Fight style combo, light punch (2x)+medium punch+strong punch+strong kick, followed by ↓ 
→ (2x)+punch. It’s just hard to do with added input lag, as I’m sure our brains remember how to precisely excute something we did hundreds of times in our lifetime. I wasn’t even planning on completing World Tour, this alone says a lot.
On Ymir, it worked mostly as I recall on original hardware, also, Akuma’s secret super is a good one to try and assess input lag.
Yes, I agree, it would be great if someone with original hardware, interested in this discussion had the means to do this, it would be appreciated and very helpful for all of us.
1 Like
When it comes to options and customisation, it far outweighs Beetle and Ymir. Though those options seem to be for advanced users.
I’d test that combo here, but I sucked even on the real hardware. The only fighting games I play are simplified ones. Have you tried all the latency mitigation features inside RetroArch itself, like Max Swapchain Images and Automatic Frame Delay? They’re pretty helpful for every core, although they might be pretty taxing on the machine.
Does it still have that option which turn meshes into real transparencies? That was a godsend, and many emulators ignore a feature like this; some of them don’t even support shaders, so you’re stuck with dithering. Besides SSF, I remember only Snes9x doing something similar (blending jailbars on hi-res games).
Wait, I think Beetle PSX can blend dithering before presenting the frame, but I can’t test right now.
I used SSF for a few years, until Retroarch and Mednafen. I’d say that in this case, more is less, many of these options in such Win95 style tabs are a nightmare in today’s age. Take PCSX2 for instance, it used plugins and the user would have to fickle around to see which options, and there were tons of them, stick. What happens today is that many fixes automatically happen under the hood, this is also true for Swanstation and a few more.
The goal is to make the software as user friendly as possible and less hacky as possible, and Shima, SSF’s developer never cared enough to modernize SSF’s UI or unifying the core emulator, it does indeed has tons of features, but it overwhelms. Did you know that Saturn Tribute on the Switch uses SSF as a base for the emulation and that is the reason why it was launched with horrendous input lag? Meaning, more than 7 frames, sometimes in the 10’s. I don’t know if all Saturn Tribute titles after the early ones still rely on Shima’s work, but I’m certain they aren’t the best representations of the originals, as far as gameplay goes.
Not only that, but the way SSF handles its .ini files, saves and more are convoluted and counter-intuitive.
As discussed about it a few years back, when Shima pulled SSF’s source code from Github as soon as City Connection contracted him to work on the Saturn Tribute releases.
It’s sad, it would had been SO much better if instead of Shima, M2 was to work on such titles. I do mean that Saturn emulation has suffered from severe input lag since the beginning, Ymir is the only that feels like it should and it doesn’t even have internal overclocking or runahead features yet.
I just uploaded this short video showing the combo on Ymir:
3 Likes
Game had already severe lag on Saturn. Steam version is a little better compared to Switch and Ps4. I was mostly referring to emulation issues as SSF has already few years advantage.
It should add a runahead or sync option like most emulators do but in its current closed source and by one developer only, it is difficult
Emulator is far too insulated,like most Japanese emulators (Tsugaru, Sharp X6800, Ootake etc)
1 Like
It’s a matter of identifying the problems. Both have similar performance, and the differences are too slight to justify saying that either one is too slow. You may have configuration issues.
I was reading the Mednafen forum, version 1.29.0 breaks several games, it seems to be related to the Extended RAM cartridge, I didn’t look into it too deeply, those errors are fixed in 1.32.1.
If that were the issue, but you’re comparing emulators, and the easiest and most accurate way to check input lag, without suffering from additional hardware output, is by advancing frames.
Edit:
The input lag of mednafen/beetle is lower than that of Ymir.
Try performing this combo in Arcade mode. I suspect this is an old issue with RetroArch.
1 Like
Differences are not slight:
a 30 fps difference tells me that they do not have similar performance.
No, but you are welcome to do some tests comparison with uncapped framerate.
I believe ymir is a bit faster than RA beetle saturn when i tested them on i7-7th gen. Ssf and yabasanshiro (too unstable, could crash anytime) could keep solid 60 fps while ymir could mostly, if you had disabled some options like sh2 cache. Beetle couldn’t except if you run some 2D, it had better compatibility than ymir though.
1 Like
I’ve noticed that Beetle-Saturn doesn’t handle multithreading the same way as standalone Mednafen, so it can be a bit laggy if your CPU doesn’t have a high clock speed for good single thread performance.
Another option to improve performance would be to integrate the dynarec from Yabause into Beetle Saturn, which I started a whole thread about. The dynarec does work, but a problem with dynarecs in general is that the assembly code can introduce some tricky dependencies on the underlying operating system and shared libraries. Some developers might consider this a ‘messy’ solution that risks breaking due to OS upgrades or changes to the build environment.
Making the emulator multithreaded risks introducing bugs due to thread synchronization issues, so developers can also be reluctant to do this.
There are various ways to improve performance, but it does come at a cost of complexity and requires more time for testing and debugging.
2 Likes
It would be very logical for Ymir to be faster than Mednafen.
Mednafen is an emulator built from scratch and is open source, while Ymir already knows how everything should be done, so it can do it faster and more efficiently.
Ymir is in the Playable phase, Mednafen aims for 100% error-free, it has been around longer, has more corrections and therefore more code, so it is logical that it is heavier. And Mednafen supports Arcade STV (not implemented in RA), which means more code.
Is “uncapping the framerate” the same as “Fast-Forward” with the space bar? It’s not a good idea to compare a core with an external emulator, as they can be completely different functions. The important thing is performance at normal speed and monitoring CPU, RAM, and even GPU consumption.
The type of processor influences the performance of the emulator. Ymir may not run well on a P4 Dual Core.
An emulator can consume more resources, but that doesn’t necessarily mean it’s slower. It may be a more accurate emulation or use more resources for better stability.
1 Like