Hacker News Reader

Very cool, using the DGX like an “AI eGPU.” I wonder if this could also benefit stuff like Stable Diffusion/WAN etc?

Yes, these models are mostly compute-bound so benefit even more from the compute on the DGX Spark.