In this video, I look at what works when running coding agents locally on Strix Halo and Radeon R9700. I cover pi and opencode running Qwen 3.6, and SWE-bench mini benchmark results:
A very interesting video, thank you. However AI chat advice says that the R9700 does not fit in a Framework Desktop case; how did you connect it up? Did you use a GPU case with its own power supply?
Hey, I think there’s some confusion, I never did or said such a thing! I have a separate server where I have the R9700. I don’t think the video ever hints to that.
Ah, thanks for the clarification. Yes, I wonder if I joined the dots lazily and got to the wrong summary.
I’ve very limited understanding of GPUs, as I’ve never been a gamer or otherwise needed something with high-end graphics. I am a fan of the repairable Framework philosophy, and I rather fancy a Desktop paired with something like this, but I wonder if this dual-box approach is a rather expensive way to approach the problem.