1080ti replacement?

[Tetras]

My current GPU is an Asus ROG Strix 1080ti OC, which apparently can pull upto 280W. I can't fit any more radiator space in the case unfortunately, so ideally what I need is a GPU that's more modern, a bit quicker (quick enough that I can play GTA6 or similar at 1440p around 100fps happily), but with a lower power draw if possible, and there must be a waterblock available for it, which it appears narrows it down a lot these days? What happened to the water cooling crowd in the last 5 years or so?

How much does it actually use on average?

Realistically, if you're aiming for upcoming AAA games at 1440p with 100+ FPS, then it'll need to be at least a 5070 or 9070, unless the game is very well optimised.

Both of those cards use around 200-250 watts on average, so I doubt it would be much different to what you have.

You COULD step down to a 5060 Ti 16GB, which is around 160-170 watts, but they hit 80 fps in TPU's 1440p testing on average and even a 5070/9070 are barely over 100. That is with the highest settings, however GTA6 is likely a long way off on PC.

I'd say (with power usage) you're always better off going for a higher-end card and tuning them down, than buying a lower-end card to run at stock.

 

[Mcnumpty2323]

"What happened to the water cooling crowd in the last 5 years or so? "

a few things
less overhead for over clocking
undervolting/and or power limiting being
the way to go
rather than increasing power and heat

the rise of cheap exceptionally good air coolers
from thermal right especially
very cheap aio also mainly thermal right

the fact that water-cooling nowadays is more about
the aesthetics means prices are getting silly
for custom watercooling
an extreme example of that would be thermal grizzly
with their new GPU block north of £500
beautiful engineering yes but not many people
going to spend that on a GPU block
but even the less extreme examples aren't exactly cheap

and probably other things that don't come to
mind at the moment
unless it's a totally new build
then of course the current ai data centers causing silly prices
on stuff means you don't want to add the cost of
a custom loop on top