y-cruncher v0.8.5 has been released *without* #Zen5 optimizations. Still waiting for final silicon. But it is July now and reviewers need time to rerun benchmarks on older systems. (especially the new BBP benchmark) Will update with Zen5 opts later. numberworld.org/y-cruncher/

We're in business now. #Zen5 However, the 9950X is turning out to be a lot more different from the 12-core ES than I had expected. So I need to investigate. Hopefully delay won't be long. x.com/Mysticial/stat…

Turns out an optimization that I did on the 12c ES, backfired catastrophically on 16c. Can't give details yet. But it made the #Zen5 binary *slower* than the Zen4 one on the 9950X. Opt has been undone. Release will be in a few hours when I'm more awake. x.com/Mysticial/stat…

#Zen5 optimizations are out for y-cruncher! Perf gains are bigger for ST than MT. BBP does not change. May be slower on Strix due to cut-down #AVX512. But I don't have one to test. numberworld.org/y-cruncher/ x.com/Mysticial/stat…

Regardless of what happens here, my "#Zen5 #AVX512 teardown" article will still go out on time. Though I may need to rewrite parts of it if it becomes apparent that the usual media outlets will not have reviews ready when embargo lifts. x.com/VideoCardz/sta…

Things just got complicated. 2 separate embargos for the different #Zen5 launch dates. I may need to rip out my #AVX512 thermal analysis because that was done on the 9950X which launches a week later. x.com/Mysticial/stat…

Has anyone noticed that #Intel's problems seem have started when they fused off #AVX512 from client. No this is not a conspiracy theory. But it's just too funny.

Ohh... look what I get to play with for a couple days. Not sure if there's enough time to add this to my #Zen5 #AVX512 write up for Wednesday.

#Zen5 desktop reviews are out! Time for another technical #AVX512 teardown. numberworld.org/blogs/2024_8_7…

Full article is out now. Sorry for the wait everyone! #Zen5 #AVX512 numberworld.org/blogs/2024_8_7… Also, did everyone else get retail samples? Every reviewer screenshot/video of their samples seems to be missing the "D" letter mark. x.com/Mysticial/stat…

On #Zen5, y-cruncher can run faster with SMT off. Might be related to the memory bottleneck. (fyi for competitive benchmarking) Version 0.8.6 will have a new option to not use the SMT cores. (lazy way to turn off SMT without going into the BIOS)

Going through my late uncle's stuff... He has a #Chromebox with a password that I don't have. Drive appears to be encrypted. Is there anyway I can get in? I have full control of his email listed on the local account. But the password for that gmail account doesn't work.

For the past half a year, I’ve told multiple high-ranking #AMD engineers to do this even if it’s just a halo product. At no point did I get the impression that this would actually happen - not even in body language. I’ll believe it when I see it. And yes I want it!

Here’s the problem with the 7950X3D: 1. Gamers hate it. (expensive, scheduling broken) 2. Devs hate it. (X3D not helpful, clock drop) 3. HPC hates it. (needs symmetry) 4. “Productivity” hates it. (scheduling broken) 5. People just hate asymmetry. A dual X3D chip would fix 3-5.

y-cruncher v0.8.6 Preview: Tuning Profiles Within each binary, you can select which profile to use. inb4 I get ripped again by the competitive benchmarking community for more perf changes...😅 numberworld.org/y-cruncher/new… (Thanks to Framework for providing the FW16!)

New y-cruncher released! I held out until CES in the (longshot) hope that AMD would release a dual X3D chip so I could add a profile for it. But of course... nice things rarely ever happen.🙁 github.com/Mysticial/y-cr… x.com/Mysticial/stat…