Upscalers often allow for a better IQ/performance balance than would be practical otherwise.
I have nothing against using them, where I determine they provide the best overall experience, but I think that designing around the assumption they'll be used is folly. Judging something negatively because of their absence, without experiencing the title first, is also premature. They are nice to have, but the implications of them, especially the proprietary incarnations, are complex, and not entirely positive.
The slowest GPU I've used DLSS on is an RTX 3080 and I still use DLSS on my watercooled, 3.05GHz core, 24.3Gbps memory, RTX 4090.
A 'lagfest' would be categorically unacceptable to me, but there are scenarios where I'll use frame generation. The frame rate that latency becomes entirely acceptable is generally lower than the threshold where motion smoothness does. So, outside of scenarios where even imperceptible levels of latency can negatively impact gameplay (e.g. competitive shooters), I'd much rather have well-over 60 fps with frame generation than sub-40 fps without it.
Cyberpunk is an interesting case, because, by default, the game has quite high latency. So much so that using frame generation doesn't even add any relative to the latency floor that existed before they added Reflex (at least on NVIDIA parts, where the low latency modes other than Reflex do not work in DX12 titles). Of course the game feels perceptibly more responsive if one just enables Reflex without frame generation, but it was quite playable before, and is still quite playable after, while making path tracing viable.
Last edited:
