2 ms·
> 3.2Tbps bandwidth PER MACHINE Looking into the Ultraclusters page you linked to in a sibling comment, it seems like the host machines pretty much fill out th
by justinclift 3y ago
> 3.2Tbps bandwidth PER MACHINE
Looking into the Ultraclusters page you linked to in a sibling comment, it seems like the host machines pretty much fill out their PCIe connections with Infiniband networking to reach that figure:
EFA is also coupled with NVIDIA GPUDirect RDMA (P5, P4d) and
NeuronLink (Trn1) to enable low-latency accelerator-to-accelerator
communication between servers with operating system bypass.
https://aws.amazon.com/ec2/ultraclusters/ https://aws.amazon.com/ec2/ultraclusters/