DriveNets Advances Open Backend Networking for AI with AMD

DriveNets works with AMD to deliver an open, full stack backend networking solution for AI advanced by AMD Instinct™ MI350 Series GPUs.

Dudy Cohen, VP of Product Marketing at DriveNets, shares how DriveNets looked for a GPU maker that is open and that will allow the company to be synergetic to it and together build an end-to-end architecture. With AMD, DriveNets finally has an end-to-end solution that is open, tested and validated.

 

Full transcript

DriveNets is a company that builds network infrastructure both for service providers and for AI builders.
So anyone that builds an AI infrastructure, we are their networking solution.

DriveNets entered this world of AI infrastructure and specifically AI backend networks a little short than 3 years ago, and we went into a market that is, let’s face it, heavily dominated by a single player. We were looking for a GPU maker that is open and that will allow us to be synergistic to it and to together build an end-to-end architecture. With AMD, we have finally an end-to-end solution, a full-stack solution, which is A, open, and B, tested and validated.

We started working with AMD’s Instinct MI350 family, which is the current technology, but we are already working and looking forward to implementing the Helios rack.

When you build an AI infrastructure, typically, you typically go to the specs of the GPU. But now when we are moving from a server scale to a rack scale and data center scale, the importance of networking becomes very clear. So when you select an infrastructure, you need to have in mind both the compute, the storage, and the networking part which connects them all. And for this, again, we have this reference architecture together with AMD that includes the whole stack and makes sure that your deployment will be fast and successful.