Lossless Scaling and emulation

[vulture]

So, guys, in case you didn't already know, there's this nice little tool:

https://store.steampowered.com/app/9...sless_Scaling/

Not only does it scale, but also generates frames up to x3 times.

Tried it yesterday with WinUAE and some well known low fps games (SF2, Outrun, Turbo Outrun, Gods) and, frankly, the improvement was borderline placebo.

Has anybody else tried it with winuae or some other 2d system emulator (ie snes, mega drive)?

How was your experience?

[mark_k]

Generating/interpolating frames isn't going to work well, because the emulated Amiga always outputs 50 (or 60) frames per second. For a game which runs at 10fps say, each set of 5 frames is identical.

[vulture]

That's why you have to set your display at 100 (or 120hz) and limit the emulator output to 50 (or 60) for FG to work and that's what I did.

And I think it did something, but as I said, borderline placebo.

I should grab with, say, obs in some lossless format with and without FG and analyze the frames to see if it really does something.

[fxgogo]

I think Mark_k is talking about the actual amount of frames the game is able to produce, not the refresh rate of the virtual hardware.

[mark_k]

Yep.

Interpolating between two identical images is pointless. If the Amiga game produces 10 different images per second, the 50fps Amiga video output will be like

imageA, imageA, imageA, imageA, imageA, imageB, imageB, imageB, imageB, imageB, imageC, imageC, ...

[vulture]

I got that, but between the final imageA and the first imageB, wouldn't the FG create an intermediate frame?

[roondar]

I suppose it would, but then that would probably only be shown for a tiny fraction of the total time the animation takes. My guess is that in that case, a 2 picture animation over 10 50Hz frames would turn into a 3 picture animation over 21 100Hz frames, with the 3rd frame on screen just 1/100th of a second.

Then again, that's just a guess on my part - I don't know how smart the software really is.

[kremiso]

wondering the results with games like Mortal Kombat 1/2, Street Fighter 2...
or games with a not short animation cycle for the main character, like Impossible Mission 2024
or Running Man (yuck)

i presume Sword of Sodan will be the same

[dreadnought]

Wow, it seems Lossless Scaling has come of age. I read about it before but it was still just a curio back then, seems it reached a "game changer" & "magic" status now.

I hope it works with old games but will now tinker and see if I can squeeze more frames from the modern ones.



EDIT: actually, I'm not sure what this github one is. Different stuff?
EDIT 2: Pirated? Nuked the link.

[vulture]

@roondar

Yes, that's right, but that's for this particular scenario. If we have a game that scrolls half frame at 50hz (25fps), wouldn't the effect of frame generation be much more profound at 100hz as a new frame would be generated 25% of the time?

Also, same scenario of half frame scrolling. If the emulator refresh is limited to 50hz(or maybe 25hz?), frame skip is set to 1 and monitor refresh is at 100hz(or maybe 50hz?), wouldn't frame generation make it full frame?

That's if the software works right of course. What do you think?

[AestheticDebris]

Originally Posted by vulture

@roondar

Yes, that's right, but that's for this particular scenario. If we have a game that scrolls half frame at 50hz (25fps), wouldn't the effect of frame generation be much more profound at 100hz as a new frame would be generated 25% of the time?

Also, same scenario of half frame scrolling. If the emulator refresh is limited to 50hz(or maybe 25hz?), frame skip is set to 1 and monitor refresh is at 100hz(or maybe 50hz?), wouldn't frame generation make it full frame?

That's if the software works right of course. What do you think?

Yo be honest, I'd expect that would make it worse if anything. A constant 25fps is reasonably smooth, but interpolating one in four frames will create a very choppy effect and that's usually much worse than a lower frame rate.

[vulture]

But what about the second scenario?

[AestheticDebris]

Originally Posted by vulture

But what about the second scenario?

I guess if your Amiga emulator is skipping every other frame (or however many is required to generate only one frame per Amiga frame) then it theoretically might work better. Would have to reconfigure everything per game though and it'd only work if the game had a constant frame rate throughout.

But I'd note they recommend using it when you're getting a steady 60FPS and have a 120Hz refresh rate. Would make an interesting experiment though, if someone wanted to try.

[vulture]

Well, I already did, but I'm not quite certain it improves the frame rate in any meaningful way.

[petran]

Retroarch's CRT Switchres on 120 hz CRT monitors yields much better results at 60 fps games, if you manage to get it running.

[Retro-Nerd]

Groovy Mister is a nice solution too (which uses Mister as a network graphics card).

[ Show youtube player ]

https://github.com/psakhis/Groovy_MiSTer

I prefer my MiSTer via HDMI & direct video mode (native resolution ouput via HDMI without an analogue adapter) though. Combined with a Retrotink 4K scaler it's a nearly perfect setup for retro games. I don't think that frame interpolation/generating will lead to something useful. At least not anytime soon.

[pixie]

Originally Posted by mark_k

Generating/interpolating frames isn't going to work well, because the emulated Amiga always outputs 50 (or 60) frames per second. For a game which runs at 10fps say, each set of 5 frames is identical.

I was under the impression that it would be 50/60 fields per second and therefore half of it in fps