It's hard to reconcile the performance across the 2 tests, perhaps they were set up/tuned differently. I wish they had published their methodology - I'd have lo

It's hard to reconcile the performance across the 2 tests, perhaps they were set up/tuned differently. I wish they had published their methodology - I'd have loved to benchmark my long-in-the-tooth RX580 & rocm-tensorflow against their numbers