Cerny said something along the lines of filling up and exploiting fully the CU rather than spread to many and have them underutilized.
It seems like the clock boost is supposedly going to help in that regard. Maybe he things the CU count with that clock boost will help push the performance better? Hence the gap will not be as big as we expect?
Also I wonder how CU count scales up, considering so many GPUs out there with different CU counts. Games wont be fully optimized with those with the highest count but with an average out there. Unless the games are intelligent enough to scale efficiently and properly
Disclaimer: not a professional graphics programmer.
Generally speaking, I think wider gpus are a little harder to feed data to, yeah, cause you need more things that you can do at once. GPUs are obviously massively parallel devices, but lots of the processing they do has to wait on some previous processing being completed. This is what makes them challengine to utilize in general -- if you only have ~50% of your compute worth of work to do right now, and all subsequent work needs to wait on that being done, half of your gpu sits empty until what you have in flight is finished and the memory is freed up. My understanding is, all else being equal, higher clock speed = less ms/ns spent waiting on each job to complete, so lack of synchronization hurts a little less.
The upside for microsoft (and every other wide gpu) is that realistically, theres a lot of independent stuff to render each frame, and there's ways to chop it up even more to make it more independent. Naturally, most of that work (big parallel computing that needs to be done super fast) is done in compute shaders these days. So ideally your gpu is maybe doing stuff like... shading some models that are ready, rasterizing/preparing others, doing big slow asynchronous calculations like GI, and maybe doing things like particle motion, vfx, etc... all at once, and getting somewhere close to maximum utilization. To ensure that, the actual cpu side code that drives setting up the pipeline and keeping stuff running can get pretty complex, with engines like Frostbite having a giant graph of dependencies and needs calculated at all times.
So summary is... there
is definitely an advantage to being faster and narrow, but in a perfect world you'd rather be fast and also wide. The ps5 will always enjoy benefits of its faster clock speed, especially when games that aren't good at filling the wide CUs are coming out, but as long as there's enough work to fill the xsx gpu that clock speed benefit isnt going to magically erase a considerable tflop/bandwidth advantage. Just shrink it.
I don't know what people expect, for example rtx 2080 has 20% advantage over rtx 2070 in current games, do you predict it will be 40% in future games ?
No -- i predict it will converge on the tflop/bandwidth advantage over time. Which is like ~25%, right? The thesis is: If a game takes advantage of both cards' architecture equally, the difference in performance will align closely to the cards' specs. This is a pretty reasonable claim!
