Yeah, read that interview. After the sixth question about variable clocks, cerny was clearly annoyed and said that coding on a translator by transistor basis is simply never even done in video games.
So it seems to me Richard just wasn't understanding the boost clocks and kept trying to compare it to PC, this im surprised by.
Then we had that RDNA 1 test he did comparing a slightly overclocked gpu vs one with a few more cus clocked lower and it didn't line up with Cerny, but it was a bad test.
We had someone post some on here ages ago that showed the kind of results cerny alluded to with a similar tf difference.
Can't remember the videos now.
With the rdna1 test he did, didn't raising the clocks beyond their limits basically pass the point where logic breaks down? They would be much better doing it at a lower clock but the same % difference within logic?
Can anyone remember the videos or a similar video of cards with a similar tflop difference showing results of the higher clocked lower tf card matching the higher tf one?
Very surprised DF would not want to show how a similar TF gpu to ps5 can actually perform similar to that of a higher tf card, unless of course it didn't fit with what they wanted to show. These are exactly the type of tests you'd expect from DF to give people an indication, but the one they did was very poor.