What are they cooking with UDNA?

>>106002869 (OP)
As an AMD user for the last 5 years, who just bought a 9070XT.

Nothing.

Whatever it is, Nvidia will outsell it 5:1 unfortunately.

>>106002869 (OP)
wait, does this fake company created by nvidia to avoid antitrust laws still exist? the charade is unnecessary now that they have more money than entire continents

>>106002869 (OP)
It will be cdna/gcn 6 + some rdna features like wmma.

>>106002869 (OP)
amd is useless shit, just go with nvidia if you need software that works at all

Besides unfucking their software stack?
I guess they can make a shader die that can utterly btfo Nvidia without being wasted on gamer only shit
Hopefully they can keep working on infinity fabric and drivers to create an in silicon return of crossfire that will let them effortlessly scale CUs by just gluing them together

>>106002869 (OP)
Mark my words, UDNA will suck massive, aids infested, gonorrhea infused, two spirited, tranny, donkey balls when it comes to gaming. There is a reason why AMD finally got competive after they moved away from GCN and created dedicated architecture for compute and gaming. AMD cannot into a hybrid architecture. It will be mediocre just like GCN was, at best.

>>106004438
>There is a reason why AMD finally got competitive after they moved away from GCN and created dedicated architecture for compute and gaming
How would you rank the RDNA generations?

There are no games I can't max out on my 9070 XT. I usually run my 9070 XT at lowest clocks + undervolted and it still runs the games I play fine at absurdly low voltages. I probably wont upgrade for years and to be honest all modern games suck these days

I think they should work on lowering the cost over increasing overall performance. Hardware has platued and the only reason to push it further is because developers are so lazy now. Stop enabling lazy developers and start making higher end hardware available to shitters

>>106002869 (OP)
their books most likely

>>106002901
well maybe amd should sell a batter product then

>>106004438
>There is a reason why AMD finally got competive after they moved away from GCN and created dedicated architecture for compute and gaming

yeah and it bit them in the ass massively, their rocm stack was wayyy behind on rdna for years, and as such they lost prosumer customers, rdna is better at somethings, but it sucked for anything non-gaming, amd has lost marketshare every year since rdna launched.

Amd's last proper GCN gpu was vega 20, it supports everything under the sun computationally, does half rate fp64, double rate fp16, and quad rate int8( which is particularly important for machine learning workloads) , rdna 1 fixes some efficiency problems like bandwidth requirements,increases total occupancy and lowers latency, but it sacrifices int8 and fp64, and there is basically no software support outside of 3d apis, you wanna know why amd had no DLSS competitior until now ? its the lack of int8, why raytracing sucks ass? rdna's compute capabilities are more limited, There was a recent patch for emulated rt on radv , and it literally brings up the rt performance so much, that the radeon VII is faster than entry level rx 6000 series cards in raytracing via brute force alone. Its frankly embarrassing, probably jeet extraordinaire raja to blame for this fuckup, only now with rdna4 have they unfucked themselves.

>>106002869 (OP)
Radeon VII 2.0 i.e. flop

>>106002869 (OP)
driver crashes from my experiences

>>106002869 (OP)
They can have an unified techstack and stop wasting so much time.

>>106002869 (OP)
negative market share

>>106002869 (OP)
a minor performance bump that's slightly worse than the RTX 60X0 equivalent but slightly more attainable