AMD has released the Radeon Pro V710 server video card, which cannot be purchased – only rented in Microsoft Azure

AMD introduced the professional graphics accelerator Radeon Pro V710, created exclusively for placement in Microsoft Azure data centers. You won’t be able to purchase the new product—its power can only be rented for computing, working with AI, and other tasks.

Image source: AMD

Radeon Pro V710 is designed for desktop-as-a-service, workstation-as-a-service, cloud gaming, and AI and machine learning workloads. The ability to use these accelerators will be available exclusively through Microsoft Azure cloud servers, where clients will be offered on a rental basis from 1/6 to 1x the power of the specified accelerator and up to 24 GB of video memory, although the video card itself received a total of 28 GB of memory.

The Radeon Pro V710 is based on the Navi 32 graphics processor with 54 execution units of the RDNA 3 graphics architecture. A similar GPU configuration is used in the Radeon RX 7700 XT gaming graphics card. However, the dedicated Radeon Pro V710 accelerator received 28 GB of GDDR6 memory instead of 12 GB, which are present in the RX 7700 XT model. The maximum memory bandwidth of the Radeon Pro V710 is stated at 448 GB/s. Radeon Pro V710 supports R-IOV technology, which allows you to isolate memory between virtual machines, allowing multiple clients to simultaneously use the same graphics card.

The card also supports hardware accelerated ray tracing, AV1, HEVC (H.265) and AVC (H.264) encoding and decoding. Additionally, it has support for AMD ROCm software. The latter, together with hardware AI accelerators for efficient matrix multiplication, provide increased computing performance for machine learning.

Image source: VideoCardz

The thickness of the video card is one expansion slot, it is equipped with a passive cooling system – the case fans of the server itself will be responsible for the airflow. Compared to the gaming Radeon RX 7700 XT, the Radeon Pro V710 has twice the memory, higher memory bandwidth, but its GPU clocks at 500 MHz than the gaming solution. In addition, the stated TDP of the dedicated accelerator is 158 W, which is 35% lower than the gaming version.