It is an AMD Radeon RX Vega 56 GPU adopted in Google Stadia

With previous news, we had the opportunity to talk about Google Stadia: this name indicates the American company's initiative aimed at making cloud gaming accessible to the various device types on the market, obviously all driven by Google.

We point out the GPU adopted in its servers dedicated to the Stadia project among the information initially provided by the American company. These are AMD-based solutions with 56 computer units and HBM2 memory.

These technical features immediately led to the Vega family's thinking that these were proposals: after all, they are combined with the HBM2 memory. Thanks to the information provided by Kronos, the organization following the development of the Vulkan API, it emerged that what was adopted by the GPUs used by Google for Stadia is an AMD GCN 1.5 architecture This acronym identifies cards based on Vega architecture, but not the latest Radeon VII equipped with Vega GPUs built with 7-nanometer manufacturing technology. The latter, in fact, has GCN 1.5.1 specifications, whereas those GCN 1.5 refer to Vega cards equipped with a 14-nanometer-built GPU.

That's why with the GPUs for Google Stadia, AMD has obtained a brilliant way to use GPUs whose development process is now quite mature, ensuring a presumably large volume order. The presence of 56 compute units suggests the Radeon RX Vega 56 card's basic specifications, although Google's reported video memory bandwidth is as high as 484 GB / s and therefore comparable to that of the Radeon RX Vega 64 card.

Therefore, specifics are obtained that are halfway between the two Vega 56 and Vega 64 cards, with presumably also changes in the frequency of the GPU clock compared to that of the Vega 56 model in order to achieve a further increase in performance compared to that already obtained from the memory side.