geordiemp
Member
It's the same thing for CPUs and GPUs, clock gating has got nothing to do with allowing for variable or higher clock speeds. They are two different concepts, as I have backed up with many different sources. Also, the shader arrays are not turned off, as it's fine clock gating, the power savings go way finer then that, to the register level.
The XSX is clocked lower then the PS5 not due to it missing any features that allow for higher clock speeds, or that the GPU can't run at those frequencies, but because it is constrained by the thermal and power limits of the consoles design. A 52 CU GPU clocked at the frequency of the PS5 GPU will draw way more power and require a larger cooling system. It's the same reason you are not seeing the PS5 pull frequencies in excess of 2.6Ghz, the GPU can likely do it, but it will blow past the allowed power limits and thermal dissipation of its cooler.
The XSX is a larger GPU then the RX 5700 XT running at similar frequencies but drawing 100w less under load. How do you think that was achieved? By perhaps having the RDNA2 architectural changes that AMD mentioned (such as the clock gating) that allows for greater than 50% performance-per-watt gains?
You cant compare against RDNA1 for starters. XSX die is 20 % bigger than ps5 so your way out estimate is not true. 22 % diff in frequencies but its not linear vs power. If your argument were true XSX would still be 2.0 Ghz or thereabouts, not 1.8 and it would be variable to switch around power where its needed.
I was being sarcastic about shader arrays turned off, that was a joke. Registers and cache are shared in a shader array, so the parallel use if very different to a CPU.