Blur Busters Forums

How do I know, if you didn't explain it in the first place? And what do you mean with "more work for the frame"? The guy asked about the input latency of an certain graphic setting, and I gave him my answer.

FPSMaster wrote: ↑

12 Sep 2021, 16:37

And what do you mean with "more work for the frame"?

He's saying that even if you aren't GPU-limited, GPU-heavy settings will still increase render time, but as I said in my previous comment in this thread, that's not what most consider or refer to as traditional "input lag," since it's part of the render time in this case, and not actually delaying delivery of an already rendered frame (which can be much more impactful).

Basically, GPU-heavy settings lower the average framerate, as each frame now takes longer to render whatever extra setting (AA, higher res shadows, more complex lighting, etc) is enabled.

I guess a clearer way to put it to the OP, is if the AA isn't significantly lowering your average framerate when enabled, it's not notably increasing your "input lag," which in this case, again, would be directly due to a lower sustained framerate.

FPSMaster wrote: ↑

12 Sep 2021, 16:37

How do I know, if you didn't explain it in the first place? And what do you mean with "more work for the frame"? The guy asked about the input latency of an certain graphic setting, and I gave him my answer.

Input lag is a long, complex chain.

Crudely simplified chain:



The input lag chain is affected by many weak links.
Here, frame rendering is done by GPU.
Frametime is one of the many sources of GPU lag.

Frame rendering time (frametime) is one of the latencies in the latency chain.
The longer it takes to render a frame = The lower the frame rate = More lag
The longer a frame waits for the monitor (e.g. VSYNC ON) = more lag
More work for frame = lower frame rate = more lag

Enabling AA lowers the frame rate, which increases lag.
Increasing resolution lowers the frame rate, which increases lag
Adding AA lowers the frame rate, which increase lag.

50fps = 1/50sec = 20 milliseconds frame rendering time per frame
100fps = 1/100sec = 10 milliseconds frame rendering time per frame
200fps = 1/200sec = 5 milliseconds frame rendering time per frame

Frame rendering time: Imagine GPU as simply a very fast electronic artist.
The GPU needs time to "paint" and "draw" a frame.
The GPU has to repeat this many times a second.
The monitor has to wait for the GPU to finish painting a masterpiece of art (one stationary picture, aka "frame")
At 100fps, the GPU is painting 100 art masterpieces per second. Monitor only needs to wait 1/100sec for each masterpiece.
You want prettier graphics? More work for the GPU. The GPU needs more time to paint/draw each frame.
You increase detail level lots. The frame rate goes down. The monitor does more waiting for the GPU.
At 50fps, the GPU is painting 50 art masterpieces per second. Monitor now needs to wait longer at 1/50sec for each masterpiece.
Prettier can be via AA, or raytracing, or extra detail, or better shadows, or all the above, or anything that adds more work per frame.
Anything that increases frame time = lower frame rate = more lag.

More work for frame = more frame time = lower frame rate = more lag
Adjusting ANY graphics detail bigger/slower (including AA) = lower frame rate = more lag
Anything that lowers frame rate = more lag.

In a nutshell, we have two example scenarios:

Cap game at 240 fps with low settings
Cap game at 240 fps with high settings

(lets say both 50% gpu usage)

Input lag is equal.

Lets say the same scenario but >80% gpu utilization, frames can start to queue yet the target framerate of 240 fps is still maintained.

Result: Potentially more Input lag, but not due to render time/longer frame times. Frames are still 4.1666 ms each.

There are multiple causes of increased input lag.