2 ms·
Assuming you have no upper bounds in terms of cost, it's probably the AMD Instinct product for simple inference and Nvidia DGX products for mixed workloads.
by bigyabai 1y ago
Assuming you have no upper bounds in terms of cost, it's probably the AMD Instinct product for simple inference and Nvidia DGX products for mixed workloads.