The graphics board includes two Vega graphics chips along with enough memory to carry out heavy-duty remote workloads for up to 32 concurrent virtual users
AMD has launched what it calls the first hardware system specifically designed for Virtual Desktop Infrastructure (VDI) in the form of the Radeon Pro V340, which the chip maker said can accommodate 32 concurrent virtualised users on a single graphics card.
The V340 features two Vega graphics chips on a single board along with 32GB of second-generation high-bandwidth memory (HBM) with Error Correcting Code (ECC), with each Vega chip handling up to 16 virtualised users.
That means each virtualised user is allocated 1GB of memory to power remotely run graphics-intensive workloads such as CAD simulations, rendering and Desktop-as-a-Service (DaaS).
AMD said its architecture allows for scalability without incurring the steep hardware licensing fees that can be associated with other arrangements.
The company compared its system’s user density favourably to Nvidia’s competing Tesla P40, which supports 24 concurrent users.
The remote desktop architecture ensures a certain amount of security, since no data is stored on the client, and the new AMD board also includes a built-in security processor providing secure boot and data encryption.
Each client is isolated from the others at the hardware level, AMD said.
The card includes an integrated encode engine that can independently compress video streams in H.264 and H.265 formats, shifting those workloads away from the CPU.
AMD announced the board at this week’s VMworld in Las Vegas, but didn’t disclose exact specifications for the GPU and provided no pricing details. The V340 is set for availability from the fourth quarter of this year, AMD said.
Nvidia recently unveiled its first high-end chips based on the next-generation Turing chip architecture, but hasn’t yet announced a Turing-based product for data centres that would supplant its Tesla cards based on the older Volta system.