points
Sorry, I meant before the M1 came out. And you and I both know that "unified memory" doesn't refer to allocating ram to the gpu for zero-swap sharing.
Vega series is 2017.
https://rocm.docs.amd.com/projects/HIP/en/docs-6.3.0/how-to/...