Since you included the GTX 1080 and 1080Ti in there you must also include the following: 2070, 2070 Super, 3060, 3060Ti, 3070 and 3070Ti, as these cards offer same or better performance than 1080/1080Ti. This will boost the percentage significantly.
NVIDIA also designs one arch and reaps the cost across AI/Professional/HPC/Gamers/Mobile products, so the much bigger R&D argument doesn't hold here.
Like I said, it's not about relative performance. The same buying decision for console and PC gamers is the high-end option or not.
100% spot on
They had the Infinity Cache option last time around too, but neither console manufacturer wanted it
There it is. Sony target customer for PS5 Pro! Sony know what they are doing.
If some rumors are true, then PS5 Pro will be a whole new GPU (well APU). Most credible ones are that it will be some kind of RDNA4 prototype for AMD, the same way PS5 was a RDNA2 prototype. The first RDNA4 desktop GPU should have the same restructured architecture, the same number of WGPs by shader engines and the same new traversal RT hardware as PS5 Pro.
Good point. I wonder if that's related to them all being RDNA1 (no IC) with a bunch of RDNA2 (all/largely with IC) features bolted on? Maybe simply cost? I'd love to know why they passed up something that PC benchmarks proves as beneficial.
So I am. And you don't need dedicated hardware to do ML jobs as it can be done easily with FP16 like in God of War Ragnarok. Spider-man games also do ML inference on PS5 but i don't know how they do it. The point is you can perfectly do it without dedicated ML hardware available since RDNA2 era. Maybe Cerny will think the FP16 compute they have is enough to do what developers need for their games. I could see them adding ML silicon (like what's in RDNA3) if they had a new need with a specific upscaler technique similar to DLSS, but I doubt it because we haven't seen a Cerny patent looking anything like a DLSS technique.
I'd argue that with cross-gen now the standard, PS5Pro would get, and inspire, higher-end PC features. Devs wouldn't target 5Pro directly, or even need to; they'd just make their cross-plat game in UE5 or whatever and set the appropriate settings, without needing bespoke 5Pro solutions like 4Pro needed.
No comparison was being made towards any conclusion. It was just an observation.
For PS5 on the hardware side, there was the custom I/O controller and weird solid state cells (slow writes, fast reads), plus whatever backwards compatibility AMD had be implement in hardware because PlayStation has well-documented super thin architecture abstraction layers. I reckon the vast majority of R&D is on the software stack and creating new tools and profilers for developers as this is produced in-house by Sony (SN Systems) and takes a long, long time - along with all the documentation ghat developers need.
Depends entirely what you are comparing it to. Notably, Dictator must have had an idea of what 4Pro should have sold that would not be incredibly low. What would a reasonably expectation be, and why? Notably because it had never been done before, what's the basis for thinking it would sell significantly more? Or what's the basis for thinking it would sell any number whatsover?
There were already about 40M PS4s before PS4Pro released. Therefore actual percentage of hardware sales was more like 14/80, or 17.5%.
It was closer to 50m. PS4 Pro launched in the second week of November 2016, three weeks later Sony passed 50m PS4 consoles.
That's more units sold than Apple sells iPhones.Of course it's arbitrary, it's a complete nonsense. I think in the UK the cost difference between the cheap PS4 Slim and PS4 Pro was £240 vs £350 but consoles and PC performance and pricing really aren't comparable because PCs and their commodity parts need to be sold for a profit and consoles do not, but that does doesn't stop platform warriors trying to do just that.
