Digital Foundry provided another look at FSR 4 and generally agree with HUB’s conclusions: YouTube

I’m still lowkey a bit confused how AMD managed to leapfrog their ML-upscaling solution from nothing to >DLSS 3 in one generation. Then again, XeSS was pretty good on Intel GPUs too hmmmm

You could probably just power limit a 9070 XT and get basically the same kind of efficiency though right? You wouldn’t even have to go through the (potentially) long process of tweaking and testing like you would with an explicit undervolt / overclock etc so it’s not exactly a tedious or difficult thing to do

It’s pretty good (so long as you have enough CPU power to be minimally affected by the higher driver overhead), but it’s in a different (lower) performance category for now. We’ll see how the B7xx stuff lands though!

The mbin equivalent (which is relevant to the OP) is More -> Open original URL or Copy original URL

Unfortunately seems to not work for mbin (which fedia.io runs)

Unfortunately YouTube is one of the only large platforms currently accepting AV1 ingest, so most places you’re still stuck with H.264, or maybe H.265 if you’re lucky :(

So… they did the opposite of the previous problem? Instead of launching too high then quickly lowering the price, they decided to launch low but then increase it (with some unknown future carve-out to make it a “business decision” instead of “lying” 🤔