Carzy Zarx’s PC Gaming Emporium - Catch Up on All the Latest PC Gaming Related News

[JEMC]

haxxiy said:

JEMC said:

I really hope those claims for UDNA/RDNA5 are legit, but when it comes to AMD GPUs, you have to take those leaks with huge amounts of salt.

After all, that's what they're trying to achieve, their goal, and they have failed before (RDNA3 was supposed to bring another 2x performance per watt jump over RDNA2 like the later did over RDNA1, but it fell short of that by quite a lot).

UDNA is CDNA-based plus it's the architecture Sony and MS have a stake on, just like RDNA2. Besides, it's just the same leap as RDNA4 vs. RDNA3.

So it's pretty beliveable especially given how much money CDNA is getting vs. the pennies the RDNA folks work with, TBH.

The fact that Sony and MSoft have gone with UDNA doesn't really say much about it, besides the fact that it's going to be AMD's only architecture going forward, making it the only choice both console makers had, besides sticking to a, by then, old RDNA4 architecture for their machines.

If you have any source that describes UDNA, to compare it to both RDNA4 and CDNA and see to which one it looks more like, I'd appreciate it.

In any case, my comment wasn't about the architecture itself, but the fact that, sometimes, what sounds good on paper or simulations, doesn't translate into real world performance. We've seen it with both AMD and Nvidia in the past, with cards that should deliver more performance than they actually did based on the specs they had, but they didn't because there was an unnoticed bottleneck that hindered the overall performance.

And then, there's also the node it will use. It shouldn't happen, but we may find that in such dense chips, the promised extra speed or lower consumption isn't quite there.