Some of us will have been monitoring for other implementations of the Unified Memory Architecture, already provided by Apple with definite and resounding, while

Some of us will have been monitoring for other implementations of the Unified Memory Architecture, already provided by Apple with definite and resounding, while naturally partial (it remains a "speed vs cost" compromise), success. The article explores both NVIDIA’s DGX Spark ("Project DIGITS") and AMD Ryzen AI MAX+ 395 (Strix Halo), and Apple M4 MAX. The above said compromise remains apparent: Nvidia: 273 GB/s, 3k$; AMD: 256 GB/s, 2k$; Apple: 546 GB/s, 5k$. (Prices may be speculative.)