There's a difference between claiming max performance and minimum. Nvidia's are actually advertising lower performance metrics instead of what their gpu's can achieve as max.
MS guarantees their performace metrics, it's the bare minimum performance you can expect. In Sony's case they advertise whats possible at the most. It's an advantage, hence they emphasise it. It's also different to what a RTX gpu does, it has a baseclock, but can boost if the situation allows it. In PS5 it's the other way around.
They couldn't even achieve 2/3Ghz for the GPU and CPU respectiively, now with boost and variable clocks they can maintain 2.23/3.5.
If it never happens (1% of the time, and when it does, it barely downclocks), one can wonder why even bother implementing it or even talking about it.
My guess is the gpu is variable between 9 and 10tf when both cpu and gpu are hammered.
I disagree with some of your words:
Tflops is a theoretical number... It relates to the bare maximum output a system could have. In other words, if you could squeze every "nut and bolt" of the system, that would be the performance you could have.
The formula to calculate that maximum is: clock speed x cu x 64*2
In this formula, where does the workload enter??? It just doesn't. Its assumed to be at maximum.
A fixed clocks system doing nothing is not outputing the same as a system with a medium workload, and even less than a system with a full workload. So output in processed Flops is not a consequence of only clock speeds
So... A fixed clocks system is not better in performance than a variable one. Clocks by themselfs do not represent performance, and performance depends on workloads, and clock speeds only need to be enough to deal with that workload in the allocated time given to process.
This means that, as far as I see it... with two equal GPUs, one with fixed clocks and other with variable clocks, there is absolutely no diference in performance terms in both systems.
And this is my first disagreent with your wording, where you claim that one system claims the possible maximium performance and the other the bare minimum performance.
If systems are equal, on maximum workload, both will have the same clockspeed, and will have the same output. So... where is the diference?
But in fact there is a diference... As you well say it!
With fixed clock speeds Sony was having problems with 2 Ghz... But now, with variable clocks, and workload control (directly related to watts consumptions), they are at 2.23!
This means that if we take the above same equal GPUs, the one with variable clock speeds would achieve higher performances than the one with Fixed clocks.
The second point I disagree is you claiming that the system varies between 9 and 10 tf... As explained above the real Tflops processed (aka output) depends on workloads, so even Xbox will have variable outputs, even with fixed frequêncies.
Otherwise, PS5 will work as Xbox séries X, only diference beeing it can achieve higher frequêncies, and clocks will not be fixed, but variable according to demand.
