Reshade, GPU and bottleneck

[hejonar]

I am trying ReShade for a certain game that also needs lowest input lag possible. Game is very CPU dependent, in fact I am CPU bottlenecked, while GPU runs at 40% usage tops. I found few ReShade scripts that I wanted to run and noticed that they all combined require like 0.01m/s of CPU and 3.00m/s of GPU, at least according to ReShade itself.

I am not sure how Input Lag works from technical level. So how would running those post-process scripts would affect my input latencies? Given I have massive GPU headroom and game is running with FPS limiter, would those extra 3m/s be still added?

Note, I have tried software called CapFrameX and according to it those 3m/s are added(22m/s->25m/s). But I've also read somewhere that Input Lag can't be measured by software and CapFrameX Input Lag measurement is purely synthetic.

[Chief Blur Buster]

It depends on how powerful your GPU is.

On one GPU, the same ReShade filter may add +10ms lag, and on a different GPU, the same ReShade value may add only +0.3ms of lag. Or maybe +20ms versus +0.1ms. Or maybe +5ms vs +2ms. It depends on the filter, depends on ReShade configuration, and depends on how fast your GPU can execute the ReShade filter. Some filters execute in only 0.1ms on a RTX 4090 GPU.

The slower your GPU, the more costly a ReShade filter is.

[hejonar]

That's very interesting, I had no idea it works this way. Thanks.

[Nocta]

The slower your GPU, the more costly a ReShade filter is.

Do we have data for CRT-Royale in the cases of the most likely cases: mid-range cards (xx60-70/ti) and APUs (SteamDeck)?

This is so far the most versatile CRT shader I tested (please chime in if there is a better alternative) and It'd be interesting to know how much input lag it's adding depending on the situation, to see if it makes sense to use it depending on the scenario.

[RealNC]

I am not sure how Input Lag works from technical level. So how would running those post-process scripts would affect my input latencies? Given I have massive GPU headroom and game is running with FPS limiter, would those extra 3m/s be still added?

These numbers simply reflect current frame time increase. Added latency is simply the result of lower FPS. But, if the GPU has frame time to spare, meaning FPS is not affected, there might not be any lag increase at all.

When not being CPU limited, and you're not using an FPS cap nor vsync, then if without reshade you're getting 200FPS for example, but with reshare you're getting 125FPS, that's a difference of 3ms (1000/125 - 1000/200 = 3). This will indeed raise your input lag by 3ms.

If you're CPU limited, or you capped your FPS, or vsync is limiting your FPS, then the GPU usually has frame time to spare, so you get the same FPS with or without reshade. In that case, input lag isn't affected as much. In this example, this simply means that if you're capped to 120FPS to begin with (due to an FPS cap or CPU-bottleneck) then the 3ms cost of reshade won't have an effect, since your 120FPS cap is lower than 125FPS.

[Nocta]

Thanks a lot RealNC for the clarification!

[Chief Blur Buster]

Some reshade filters literally excute in less than 0.1ms! YMMV depending on the filters you choose.