I have two 128gb Strix Halos and have been extremely excited about Antirez's (Redis author) work on DS4, especially with 4bit quant using two machines: https://github.com/antirez/ds4
Right now the speed isn't good for GLM 5.2, Deepseek V4 Flash speed is okay for me (actually reading the output) and quite usable. See kyuz0's great recent video here: https://www.youtube.com/watch?v=PkKXm_mKCCM
With a bit more speed and model improvements, local AI becomes a reasonable practical thing! The biggest problem is all the tech companies making consumer hardware completely unaffordable, and I don't think this is accidental. Look at Micron's profits and share price lately...
I got my Strix machines for ~2k eur each, best computers this 90s kid has ever owned, but those days are gone :(
This memory bandwidth combo is amazing for homelabbers. kyuz0's work on these containers has made the investment in this kit so valuable I hope Framework is sending you hardware!
https://projectbluefin.io/server/ is what I'm hoping to ship, designed to just ship setups like this ootb and things like this would be so much harder without kyuz0!
(Note: The 64GB ones are going for $1700-ish empty, the prices on the 128's are outrageous we can just keep making the labs more deterministic over time!)
I have two 128gb Strix Halos and have been extremely excited about Antirez's (Redis author) work on DS4, especially with 4bit quant using two machines: https://github.com/antirez/ds4
Right now the speed isn't good for GLM 5.2, Deepseek V4 Flash speed is okay for me (actually reading the output) and quite usable. See kyuz0's great recent video here: https://www.youtube.com/watch?v=PkKXm_mKCCM
With a bit more speed and model improvements, local AI becomes a reasonable practical thing! The biggest problem is all the tech companies making consumer hardware completely unaffordable, and I don't think this is accidental. Look at Micron's profits and share price lately...
I got my Strix machines for ~2k eur each, best computers this 90s kid has ever owned, but those days are gone :(
What's the advantage of ds4 over llama.cpp, esp if down the line they upstream his forked kernels?
This is amazing!
I'm working on a three node strix halo agentic OS factory designed to be maintained by local agents: https://github.com/projectbluefin/testing-lab
This memory bandwidth combo is amazing for homelabbers. kyuz0's work on these containers has made the investment in this kit so valuable I hope Framework is sending you hardware!
https://projectbluefin.io/server/ is what I'm hoping to ship, designed to just ship setups like this ootb and things like this would be so much harder without kyuz0!
(Note: The 64GB ones are going for $1700-ish empty, the prices on the 128's are outrageous we can just keep making the labs more deterministic over time!)