No. There is already a 520mm2 GPU on a 7nm process with 40WGs running at clocks higher than that in the PS5. (6950XT). ANd even higher costs and more WGs on a 5nm process if we go into 7900XT territory.
There really s no reason why Sony would have to keep double its CU count to maintain compatibility. That can be sorted out with a better-designed tool kit. I personally have never believed that whole double CU count to maintain compatibility thing. It can be abstracted.
I don't see them doing anything with the RAM outside using faster chips. I can see them increasing the cache within the chip though. A PS5pro with 64MB of infinity cache on the chip would do more, and be cheaper for them than going for 24GB of RAM and a 384bit bus (which is what you would be getting if you are using 12 chips instead of 8 as hitting any other number would require they adopt the convoluted memory system of the Xbox.
I believe that's all still targeted, going by the OP. 8K as an option won't diminish 4K, it'll likely improve it further. Remember when the Unreal demo showed 8K textures in 1440P res?
While I agree, new hardware won't impact that situation one way or the other.
From a cost/power/heat perspective it makes sense to do what they are supposedly doing I think.
My bad read too fast and didn't see the pro slim in all honesty and yeah that doesn't make any sense
Leaks will pick up once more 3rd parties get the units
This assumes the launch PS5 and the PS5 Pro SoCs would use the same process node, and they obviously won't.
The "butterfly arrangement" of the PS4 Pro was less than ideal. The chip had more pixel fillrate than its memory bandwidth could ever allow utilizing, i.e. a bunch of transistors / die area dedicated to ROPs that never got any utilization.
Aww, you've got your cocky tude on...bless.
do you feel good now? There's a good lad.
Chiplets on Consoles don't make sense, you want everything in a SOC(system on a chip).
It will use RDNA 3 and RDNA 4 features
We're already getting chiplets on sub-$450 graphics cards. This Pro console is most probably going to cost over $600.
Still not enough to do acceptable RT at the resolution, frame rate, and fidelity people expect.
Problem here is does Tsmc have the Capacity for it? RDNA3 uses 2.5D fanouts for it Packaging and these things are limited by capacity.
The 7600 is on 6nm not 5. RT on rdna 3 is part of the cu, so the gain in rt is the same as the gain in cu IPC. That's why the 7900xtx still has crappy RT, most of the gain in rt is because it has more cus, not because they are THAT much faster at rt individually. IPC gain on rdna 3 was honestly really disappointing. It's possible that could be improved further for this ps5 pro but nobody can really say for sure until we know more.
Kepler is assuming RDNA4's RT engines, not RDNA3's.
People with 4090 GPUs aren't running games at 4K/RT/60fps. You need to turn on DLSS to get it. This article was posted before even more demanding games came out. And nobody thinks that a PS5 Pro will be as powerful as a 4090.
Honestly I think that's more a factor of diminishing returns than the consoles not being pushed.
I've never heard of the RDNA3 fanouts posing a capacity / scale problem. I also don't know if we're sure of the PS5 Pro using the same packaging as RDNA3.
Sony patent was Accelerated RT and AMD patent month ago was Traversal RT engine. That why assuming RDNA4 RT
People will buy it and its not targeting the casual, 21 voters in here and 62% say for sure
