AMD Radeon PRO V710 GPU Comes Equipped With Navi 32 Chip: 54 “RDNA 3” CUs, 28 GB VRAM & Available At Microsoft Azure
AMD has introduced its Radeon PRO V710 GPU which is designed for the visual cloud segment such as Microsoft Azure & comes with 28 GB VRAM.
Microsoft & AMD Offer The Radeon PRO V710 "RDNA 3" GPU Solution In Azure Cloud Services
Backed by the ROCm 6 ecosystem, the AMD Radeon PRO V710 is designed for a range of workloads such as AI inferencing, VDI & Remote Workstations, and Cloud Gaming. These GPUs along with the AMD Genoa CPUs will be part of the Microsoft NVads V710 v5-series of virtual machines which will feature the following host specifications:
Related Story Samsung Purchases $20 Million Worth of AMD MI300X Data Center GPUs for AI Development
Part QuantityCount Units Specs
SKU ID, Performance Units, etc. Processor 4 - 28 vCPUs AMD EPYC 9374 F (Genoa) [x86-64] Memory 16 - 160 GiB Local Storage 2 - 14 Disks 256 - 1536 GiB
IOPS (RR)
MBps (RR) Remote Storage 2 - 14 Disks 7600 - 5320 IOPS
180 - 1260 MBps Network 2 - 8 NICs 3300 - 20000 Mbps Accelerators None
As for the specifications on offer by the AMD Radeon PRO V710 GPU, it will be based on the same Navi 32 GPU as the RX 7700 XT with 54 compute units, 3456 stream processors, a 2 GHz peak clock, and up to 55.3 TFLOPs of FP16 and 27.65 TFLOPs of FP32 performance. The GPU will feature higher 28 GB GDDR6 VRAM which will operate across a 224-bit memory interface versus the 12 GB & 192-bit memory configuration of the 7700 XT and additionally feature up to 448 GB/s of bandwidth. The GPU will also come equipped with 54 MB of Infinity Cache.
As for the design, AMD has gone with a white color scheme on the passive-shroud which has the specs mentioned on the front. The card receives power through a single 8-pin connector and has a TBP of 158W. The card measures 10.5" or 267mm in length and comes in a single-slot design which makes it great for multi-instance use.
Following are the instance configurations you can get for the V710 at Microsoft Azure (for more information, you can head over here):