I would assume a lot of it comes down to power budget per core like the 64 core, but holy cow that total wattage. “Monster” results, indeed.
I’m actually curious how much - or little - benchmarks would be affected by actually running multiple instances simultaneously, because that’s kinda the whole gig.
Can’t wait to see the 32/64 PRO compared to non PRO. I think I will not be getting the 96 core because the single core benchmarks are usually less than all other models.
Those processors are insane! Love it! I will be happy with my 3995WX for a while, but when those processors are a couple years old and I can get them used, I might upgrade with one like that.
I honestly don’t feel like that Geekbench6 scales particularly well. I was unable to find information about it, but their scores are not distributed in a linear fashion. My 64 core processor scores not that much higher than a slightly more modern 16 core processor.
You asked for some ideas for testing, here are some
UE5 main-dev source compilation (and shader compilation if possible) would be great. I personally have thought of getting a system with at least 512GB RAM so that I can have UE5 source in a RAM-Drive while compiling, if that would be beneficial (sad that Optane is gone, low latency seems to be key in compiling)
On that note, instead of Optane I have also thought about CXL memory, is there a slight chance we could se a video around that in system like this?