I regret building this $3000 Pi AI cluster

I thought the conclusion should have been obvious: A cluster of Raspberry Pi units is an expensive nerd indulgence for fun, not an actual pathway to high performance compute. I don’t know if anyone building a Pi cluster actually goes into it thinking it’s going to be a cost effective endeavor, do they? Maybe this is just YouTube-style headline writing spilling over to the blog for the clicks.

If your goal is to play with or learn on a cluster of Linux machines, the cost effective way to do it is to buy a desktop consumer CPU, install a hypervisor, and create a lot of VMs. It’s not as satisfying as plugging cables into different Raspberry Pi units and connecting them all together if that’s your thing, but once you’re in the terminal the desktop CPU, RAM, and flexibility of the system will be appreciated.

I did some calculations on this. Procuring a Mac Studio with the latest Mx Ultra processor and maxing out the memory seems to be the most cost effective way to break into 100b+ parameter model space.

Depends on how heavy one wants to go with the quants (for Q6-Q4 the AMD Ryzen AI MAX chips seem better/cheaper way to get started).

Also the Mac Studio is a bit hampered by its low compute-power, meaning you really can't use a 100b+ dense model, only MoE feasibly without getting multi minute prompt-processing times (assuming 500+ tokens etc.)

Huh? My maxed out Mac Studio gets 60-100 tokens per second on 120B models, with latency on the order of 2 seconds.

It was expensive, but slow it is not for small queries.

Now, if I want to bump the context window to something huge, it does take 10-20 seconds to respond for agent tasks, but it’s only 2-3x slower than paid cloud models, in my experience.

Still a little annoying, and the models aren’t as good, but the gap isn’t nearly as big as you imply, at least for me.