Connectivity is the critical component that brings data to and from the storage devices to the accelerators, mostly GPUs. Large workloads require significant GPUs to work together to build models. A low-latency network is necessary for the GPUs to exchange data efficiently. Infiniband proclaims to deliver the lowest latencies in networking.
What about Ethernet? Ethernet can certainly deliver similar speeds to Infiniband. The Ultra Ethernet Consortium aims to provide an Ethernet-based open, interoperable, high-performance, full-communications stack architecture to meet the growing network demands of AI & HPC at scale.
In this podcast, our guest is J Metz, the Chair of the Ultra Ethernet Consortium and SNIA and a Technical Director at AMD.
As a co-host, I am excited to engage with enterprises and discuss AI use cases.
Listen in and join the conversation.