I assume a deal like that involves a lot of direct support and optimization by ROCm team for the workload the supercomputer is running, rather than generic awes

I assume a deal like that involves a lot of direct support and optimization by ROCm team for the workload the supercomputer is running, rather than generic awesomeness at everything. That’s been my experience with most large enterprise software, really. Amazing at its most critical competency, core dumps when I try something a little unusual.