Does Nvidia FREESTYLE add input lag if it doesn't drop my FPS? EDIT: yes it does

[Chief Blur Buster]

One thing mentioned,with some prosettings is "image sharpening in the NVIDIA control panel to be the best combination for competitive play.".

Any thoughts on this or straight a long Chief's post above? I recall seeing somewhere there was in fact a reason you wouldn't use the overlay for the sharpening and it should be done at the control panel.

Classic sharpen filters don't require lookahead buffer processing, so it's a processing delay without framebuffer delay. Generic shader-based sharpen algorithms are really fast on modern GPUs, so framerates should drop only slightly, and the input lag is simply the frametime difference of the two framerates (before/after). AKA, likely far less than a millisecond.

If there's an AI-based algorithm (like DLSS) to the sharpening algorithms, the processing delay may be slightly higher, but the good news is that DLSS compensates by using a lower resolution generating a higher frame rate, which doesn't slow down as much.

BTW, on the topic of sharpening -- DLSS is sort of an AI-driven sharpening algorithm (of sorts). So DLSS framerate-improvement lag-reduction outweighs DLSS processing lag -- the lag increase is correspondingly cancelled-out by lag decreases. So if you want a way of sharpening images without much lag, you can use DLSS methods at higher-quality settings.

I also write about DLSS as part of Frame Rate Amplification Technologies (FRAT). The best FRAT either maintain lag (add extra frames without increasing lag) or decrease lag (reduce frametimes). This is superior to old-fashioned TV interpolation.

[hleV]

on modern GPUs

What is considered a modern GPU nowadays?

[Chief Blur Buster]

on modern GPUs

What is considered a modern GPU nowadays?

1000-series and newer for me.

The current performance class of roughly a GeForce 1080 Ti or a 1660, can do about 5,000 to 10,000 framebuffer copy operations per second (e.g. copying the same full screen texture 10,000 times a second) in a simple C# application written in the open source MonoGame engine.

For 1080p or 1440p framebuffers, that requires at least a hundred megabytes per second of graphics memory bandwidth to bit-blit that many times per second. On modern GPUs, rudimentary shader/texture processing executes at roughly a similar speed (e.g. a framebuffer brighten / simple sharpen algorithm / contrast enhance / shadowboost / scaling / etc). 10,000 framebuffer processed per second is 0.1 milliseconds (100 microseconds).

Now, if you layer a LOT of processing stuff or layers, it's possible to add quite a lot of milliseconds of latency -- and really bog things down. But a simple single simple filter isn't going to meaningfully affect things much. Especially if you've now leaped from 50% GPU utilization to 100% GPU utilization, and now you've got some latencies associated with that too.

Your benchmark will be the lag differences of frametimes, as well as your GPU % utilization. Most of the time, your GPU is already almost 100% utilized in newer modern games when playing a game in VSYNC OFF uncapped framerates. So in this case, you're not increasing utilization, but slightly decreasing framerate, and your lag differential will be the frametime differentials.

Some filters will be massively more complex (e.g. 100x or 1000x more complex) than other filters -- so experiment with different filters to see how they affect your framerate. If your GPU utiliation increases only a tiny bit and your framerate drops a tiny bit -- then that is a fairly tiny impact on latency.

[MaxTendency]

See....Not a full frame lag. It's tinier than that.

I see, thanks for the correction as well as the detailed explanation. So there is input lag but not for the reason i was thinking.

[joe7dust]

That's already adding 1 frame worth of delay best case scenario.

Good News: Incorrect. ...
something, something I'm a tearline slaying genius

I read so much and watched several videos, but I'm still unsure what the takeaway should be for me here... I usually keep vsync off since at 144hz and high fps it is rare for me to perceive tearlines. It's not usually noticeable for me these days, definitely not like it was with <60 fps on 60hz.

In the rare cases where I do notice them and it bothers me I just turn on Gsync and be done with it. However, my display has 4ms of lag normally and Gsync brings that up to 8ms so when I saw those claims in the very beginning about being able to cleverly "code" the tearing away I was intrigued. I'm still not sure how this knowledge benefits me and what I should take away from all this... does this mean that manufacturers should have an easier time of combating tearlines in the future and I should just look forward to some kind of firmware that implements your method or is there something I can do right now?

If I could use this to cut my latency in half that would be great, but I'm not sure if that is the case here. Either way I'm glad to see such passion here, not sure I've ever seen someone so proud of a project.

[CSGO_tester]

2021

YES IT DOES!!! My fps dropped from 300 to 180. But it actually depends to the adjustment u made.