Thanks for this article. Do you know whether AMD's latest APUs (Strix Halo or Strix Point) are able to do direct memory transfers on-chip between CPU cache and GPU memory? My understanding of your article is that as of 2022, transfers had to go out to main memory (slower), but Apple chips could direct-transfer (faster), and so I am hoping that AMD has added this feature in their upcoming and recently released chips.
Thanks for this article. Do you know whether AMD's latest APUs (Strix Halo or Strix Point) are able to do direct memory transfers on-chip between CPU cache and GPU memory? My understanding of your article is that as of 2022, transfers had to go out to main memory (slower), but Apple chips could direct-transfer (faster), and so I am hoping that AMD has added this feature in their upcoming and recently released chips.