@MiaAI_lab I have two of each. Love both platforms and each have their strengths.
Strix allows me to dual boot from linux into proxmox to secondary m2 drive if I want to run VMs with the 16 cores/32 threads of Zen5 x86_64.
Spark is purely AI focused and ARM64, and that's fine, too.
@MiaAI_lab@NVIDIAAI@UnslothAI@RedHat_AI Something is off there
Same benchmark, vllm 0.24, no MTP, NVIDIA NVFP4:
35B-A3B, dual 3090: 86/100
27B, dual 5080: 84/100
@sudoingX Rocm here. And stop assuming there's only 96GB addressable - look into amdgpu.gttsize and ttm.pages_limit kernel args. I save 6GB for Fedora and podman, and have 122GB available for llama.cpp/vllm. Set your VRAM to .5GB in the bios and let Linux do the rest.
@PurpleSuede22@LunaticsBTC love it.
my question is what kind of lunacy does @TheArtistBTC have planned for backgrounds like these which look like some type of "reveal?" ๐
Aliens of Fairness is the type of collection we all should support. Not only did the founder sell them for free, but he also went through the effort of acquiring a lot of uncommons for the holders.
Respect @daimondexe ! We need more collections like this one in ordinals!
@LostatHome106@daimondexe@PipeAliens There was one more I really wanted but it went before funds tx, that's alright though.
I am happy with this awesome armored alien with 2011 sat! ๐ฅ