You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
I'm new to LLM, recently I had an idea to build a PC to run DeepSeek at home, I read a bunch of threads on Reddit and was about to get a 64 core EPYC Milan + 1024GB DDR-3200 to run the widely used llama.cpp on it, but I came across ktransofrmers.
Now I'm thinking about going with a 56-core Xeon 9480 + ~512GB DDR-4800 + RTX 3090 instead of the quiet old EPYC setup. It's a bit more expensive, but with ktransformers it seems like a better option since the CPU has AMX units, 64GB on-die HBM memory, and 24GB of VRAM on the GPU.
Unfortunately, I couldn't find any information about the 9480 CPU in the context of using it with ktransformers. There are some reports of using it on llama.cpp, but the numbers look disappointing.
I would appreciate it if you could tell me if this would really be the best option for a home DeepSeek inference setup.
reacted with thumbs up emoji reacted with thumbs down emoji reacted with laugh emoji reacted with hooray emoji reacted with confused emoji reacted with heart emoji reacted with rocket emoji reacted with eyes emoji
-
Hi all,
I'm new to LLM, recently I had an idea to build a PC to run DeepSeek at home, I read a bunch of threads on Reddit and was about to get a 64 core EPYC Milan + 1024GB DDR-3200 to run the widely used llama.cpp on it, but I came across ktransofrmers.
Now I'm thinking about going with a 56-core Xeon 9480 + ~512GB DDR-4800 + RTX 3090 instead of the quiet old EPYC setup. It's a bit more expensive, but with ktransformers it seems like a better option since the CPU has AMX units, 64GB on-die HBM memory, and 24GB of VRAM on the GPU.
Unfortunately, I couldn't find any information about the 9480 CPU in the context of using it with ktransformers. There are some reports of using it on llama.cpp, but the numbers look disappointing.
I would appreciate it if you could tell me if this would really be the best option for a home DeepSeek inference setup.
Beta Was this translation helpful? Give feedback.
All reactions