Theres evidence to believe we might be under utilizing the CPU on our EC2s as well as being bottlenecked by disk I/O and network I/O.
So, it might be worth looking into how we can speed that up similar to what we did with bf. Basically, spinning up a single GPU instance, and running various changes and seeing how it affects throughput.
Theres evidence to believe we might be under utilizing the CPU on our EC2s as well as being bottlenecked by disk I/O and network I/O.
So, it might be worth looking into how we can speed that up similar to what we did with
bf. Basically, spinning up a single GPU instance, and running various changes and seeing how it affects throughput.