Hello everyone, I have a 6600XT and I am just recently approaching these technologies. Everything I read on the web and on yt seems to contradict itself.
This is what I understand:
- FSR (and Super res): it’s upscaling and you can use it to make a game smoother (so technically I could go from 30 to 60 fps)?
- AFMF is a frame generator so technically it’s a win-more-condition, if I have 60 fps I can do better, but it doesn’t improve something that isn’t already fluid.
That said beyond the fact that FSR 2 (on Baldur’s Gate 3 for example) makes everything a bit grainy and blurry compared to native resolution, I did this test:
BG3 on an ultrawide screen at 2100x900 high details, native runs me at 43 fps.
FSR2 (quality) + AFMF: I get 190fps
But 190fps goes beyond the 75Hz of my Freesync monitor. And the frame cap in game seems to turn off. Radeon Chill cannot be enabled, so is there a workaround? Other than using custom drivers like R.ID, which seem to be stuck at the March release?
But it really makes sense for soft pacing single player to have all that FPS without a frame cap? Do I understand well the case-scenario where I should use one or the other tech?
Thanks
It’s confusing because both AMD and Nvidia call both frame gen and upscaling as the same thing.
Upscaling: GPU renders game at low resolution (eg, 720p), and then (semi) smartly guesses what’s in the pixels that weren’t rendered. You get improved framerates because the GPU is doing less work per frame. The downside is typically that the image is typically a bit blurrier, and depending on how the GPU guesses the missing pixels, you might also get ghosting, which is where moving objects leave a smear trail behind them. The general consensus is that if you plan to use an upscaler, you should only use the highest quality mode on the upscaler. Any lower and the blurring becomes too significant
Use when:
Do not use when:
Frame gen: GPU renders a frame, holds on to the frame, renders the next frame, and then guesses at what happened between the two frames. The framerate is improved because the GPU is inserting an entirely guessed frame in between every rendered frame. The downside is that because the GPU has to hold on to a frame, the latency is increased. More specifically, the time between when you move your mouse and when your camera moves will be increased with frame gen.
Use when:
Do not use when:
Terminology:
AMD FSR 1: semi-dumb upscaler
AMD RSR: literally just FSR 1
AMD FSR 2: semi-smart upscaler
AMD FSR 3: very slightly smarter upscaler than FSR 2, and comes with semi-smart frame generation
AMD AFMF: literally just the frame generation part of FSR 3, but slightly dumber
nVidia DLSS 1: semi-dumb upscaler
nVidia NSR: literally just DLSS 1
nVidia DLSS 2: semi-smart upscaler
nVidia DLSS 3: smarter upscaler than DLSS 3, and comes with semi-smart frame generation
Intel XeSS: semi-smart upscaler
This answer is so so superberb, thank you so much for clarify me this, also with the list of terminology. I didn’t know that FSR3 is an upscaler + semi frame gen… It’s a pity that BG3 doesn’t support this…
That’s a really useful answer. Thank you!