both give you ~128GB of unified memory for local AI, and their memory bandwidth is almost the same — 273 vs 256 GB/s. so the $2,700 gap buys…
The number everyone checks — and why it barely moves
Where the 2.4x actually goes: compute and CUDA
Aliteq
Nvidia's DGX Spark costs 2.4x a Strix Halo mini-PC. it barely beats it on the number everyone checks