Would PureXP potentially work on OLED displays?

[Nocta]

Actually the higher the frame rate amplification factor is, the lower the input lag gets. Lossless Scaling FG x4 has less input lag than x2. If you think about it, it makes sense. The more frames you generate, the sooner the first of those frames needs to be shown.

LSFG always buffers one frame, regardless of amplification factor. At x2, added input lag is one and a half a frame. For a 60FPS base, that's 25ms of lag increase. At x3, input lag is one plus one third of a frame and at x4 it's one plus one quarter of a frame. If there was something like x16 to convert 60FPS to 960FPS, added input lag would be one plus 1/16 of a frame (which is just 17.7ms.)

So it's not that bad

I wonder how much of a tradeoff is acceptable for fast gameplay.
I just tried LSFG x4 for SF2X on Fightcade where reaction time is crucial but motion blur can also be annoying on LCD because of the 60fps cap on this era of game. By your calculation, this should add 20ms of input lag. It's not nothing but then how much faster can you react with higher visual clarity?
I barely feel the lag but I do feel I gain quite a bit of motion clarity...

It's a tricky question I find because it's hard to be objective when you test it yourself, it would be super interesting to haveresearch paper covering this question. So far the only one I could find seems to indicate this input lag is the main factor but I don't have access to the full paper to see how they measured everything (link to the paper).

[Chief Blur Buster]

Framegen + OLED looks really nice, but most people don't want to use framegen on retro games because it requires laggy methods of framegen (interpolation). Lagless framegen is very hard to do with retro content because of various law of physics reasons.

That being said... The lag chain includes human reaction time, and most reviewers overlook the role of monitor technologies in improving human reaction time.

Improvements to motion clarity can speed up human reaction time to a magnitude faster than the latency increase by the motion blur reduction method. But YMMV, depending on the content -- some content doesn't have human reaction time gains from motion blur reduction.

There are low-lag methods of excellent interpolation (e.g. SmoothVideoProject + Elgato + RIFE 4.6+ interpolation that Elexor did) that actually worked pretty well for retro content, though it's a compromise. Once easier low-lag 4:1 interpolation becomes available, it may become a contender.

The faster frame transport of 240Hz helps quite a bit in compensating, but interpolation buffers the previous and next frame, so you have a mandatory 1-frame latency for lookbehind+lookahead bidirectional requirement of interpolation.

That being said, not all forms of framegen is interpolation, but the catch is retro content is very hard to framegen without interpolation-variants of framegen. (Framegen includes multiple spatial and temporal methods, and temporal methods that includes Interpolation, Extrapolation, Reprojection -- www.blurbusters.com/framegen)

[Nocta]

There are low-lag methods of excellent interpolation (e.g. SmoothVideoProject + Elgato + RIFE 4.6+ interpolation that Elexor did) that actually worked pretty well for retro content, though it's a compromise.

Can you point me towards more info please? I know SVP and RIFE but I couldn’t find anything about Elecard (in this context) and Elexor.

[Chief Blur Buster]

There are low-lag methods of excellent interpolation (e.g. SmoothVideoProject + Elgato + RIFE 4.6+ interpolation that Elexor did) that actually worked pretty well for retro content, though it's a compromise.

Can you point me towards more info please? I know SVP and RIFE but I couldn’t find anything about Elecard (in this context) and Elexor.

EDIT: I meant Elgato
Elexor is a forum member here who did this, and Elgato is a 4K capture card, since SVP doesn't (yet) work at the desktop level.

Possibly, a special modification to https://github.com/MolotovCherry/virtual-display-rs to pipe the current desktop through SVP before displaying, may be needed, in order to bypass an Elgato, for running local PC retro emulators through a SVP filter. But a good software developer would need to do this.

[elexor]

There are low-lag methods of excellent interpolation (e.g. SmoothVideoProject + Elecard + RIFE 4.6+ interpolation that Elexor did) that actually worked pretty well for retro content, though it's a compromise.

Can you point me towards more info please? I know SVP and RIFE but I couldn’t find anything about Elecard (in this context) and Elexor.

EDIT: I meant Elgato
Elexor is a forum member here who did this, and Elgato is a 4K capture card, since SVP doesn't (yet) work at the desktop level.

Possibly, a special modification to https://github.com/MolotovCherry/virtual-display-rs to pipe the current desktop through SVP before displaying, may be needed, in order to bypass an Elgato, for running local PC retro emulators through a SVP filter. But a good software developer would need to do this.

I think the best method right now is using the program Lossless Scaling on Steam (LSFG). It can perform up to 4x interpolation and is much less intensive than RIFE. LSFG utilizes Windows' DXGI desktop duplication to capture most games and can output with VRR support. The latency can be much better than a capture card application, depending on how intensive the game is to render and the base frame rate, of course.

The quality of the interpolation isn't quite as good as RIFE, but it can still be decent. The results are highly dependent on the type of game and the base frame rate.

[Nocta]

Thanks guys.
Lossless Scaling is indeed fantastic. X4 interpolation on 60fps emulators looks superb and I wonder if we could even go x8 for 480fps silky smooth output at one point!

I have been reading posts here and there about reducing the amount of lag created by its interpolation and what seems to come up often is using Vsync Off (in LS settings) + DXGI Swapchain (in Nvidia parameter) + Reflex On (via RTSS).

I’d love to see actual numbers when combining all of the above. Is there a reliable way to measure the induced lag?

[RealNC]

Thanks guys.
Lossless Scaling is indeed fantastic. X4 interpolation on 60fps emulators looks superb and I wonder if we could even go x8 for 480fps silky smooth output at one point!

I don't know if it's possible quality-wise, but performance-wise it's already stressing the GPU quite a lot at x4. In fact, in the games I tried it so far with, LSFG 240FPS (from 60) uses more GPU power than rendering native 240FPS

Using the "performance" version of FG makes it use about the same GPU power as native 240FPS. It's much more intensive on the GPU compared to FSR FG and DLSS FG. Which probably explains why it looks so good...

I have been reading posts here and there about reducing the amount of lag created by its interpolation and what seems to come up often is using Vsync Off (in LS settings) + DXGI Swapchain (in Nvidia parameter) + Reflex On (via RTSS).

Vsync off in LS obviously reduces input lag, but the other things are just placebo. They make no difference whatsoever. RTSS is even able to to inject anything into LS, and DXGI Swapchain also doesn't do anything because LS is neither an OpenGL nor a Vulkan app. It uses DirectX and creates a DXGI swapchain on its own anyway.

There's a lot of nonsense "tweaks" out there (and in here as well )

[Nocta]

Oh true I did not think about the impact on GPU so much because I mostly use it on a 30 year old game!

Thanks a lot for debunking the settings, it’s always complicated to know what makes sense and what is placebo without data / deep knowledge.

[Chief Blur Buster]

Vsync off in LS obviously reduces input lag, but the other things are just placebo. They make no difference whatsoever. RTSS is even able to to inject anything into LS, and DXGI Swapchain also doesn't do anything because LS is neither an OpenGL nor a Vulkan app. It uses DirectX and creates a DXGI swapchain on its own anyway.

When using LS as a strobeless motion blur reduction technology for 60fps content on PC, you ideally want to use framerate=Hz.

Using VRR+LS is best since you don't have tearing to contend with.

Making retro games look like native 240fps while still having low latency, is astoundingly difficult! Fortunately, retro games are usually low resolution.

Strobe-based blur reduction (CRT, impulsing, flicker, BFI, strobing) doesn't require more frame rate.

Whereas strobeless blur reduction via brute framerate (framegen, interpolation, extrapolation, reprojection, more GPU, etc), but you say goodbye to the flicker! Sometimes that's worth the cost of extra latency, for specific games.

[Nocta]

Strobe-based blur reduction (CRT, impulsing, flicker, BFI, strobing) doesn't require more frame rate.

Whereas strobeless blur reduction via brute framerate (framegen, interpolation, extrapolation, reprojection, more GPU, etc), but you say goodbye to the flicker! Sometimes that's worth the cost of extra latency, for specific games.

Extra latency is something one can probably learn to adapt to, no? I’d love to do strobe based reduction for fast paced games where latency matter but my screen does not support 60fps strobing (most don’t) and new shiny oled panels don’t do it either so far, so I feel like we are stuck with framegen for the moment…