Nothing yet. Waiting on things to arrive and just endless nonsense from lies to customs problems. Hope is 2xdgx sparks for big models and a workhorse with 2 x RTX 5090. It's way over the top for most ...
Check out 0xsero's stuff on running deepseek flash on 2 Sparks. It's not perfect but quite powerful. He's now running 4 Sparks but his handbook covers mostly 1-2x, iirc.
Prep some bring-up documentation in a repo with an agent for both pieces, write down what you want to achieve, which performance targets you have. Qwen3.8 Flash would work great on those RTXs I assume.
I stumbled upon some interesting Nvidia-stuff that's not applicable to my AMD stuff. Can drop you some links to lookup w/o any promises.
Prep some bring-up documentation in a repo with an agent for both pieces, write down what you want to achieve, which performance targets you have. Qwen3.8 Flash would work great on those RTXs I assume.
I stumbled upon some interesting Nvidia-stuff that's not applicable to my AMD stuff. Can drop you some links to lookup w/o any promises.