TensorFlow-DirectML support dramatically accelerates AMD adapters

AMD said that its Radeon RX 6000 graphics cards now demonstrate significantly higher performance in a number of tasks. This is GPU-accelerated machine learning based on Microsoft's TensorFlow-DirectML technology. Support for this technology appeared in the Radeon Software Adrenalin 21.8.2 driver some time ago, but there is no such data in the driver description, and AMD itself only spoke about it now.

As an example, the company cites AI Benchmark Alpha tests, where the Radeon RX 6600 XT and RX 6900 XT demonstrate a performance increase of 3.1–4.4 times in comparison with the old driver without the support of the mentioned technology. For some reason, AMD only mentions these two video cards, although other models in the line must also support Microsoft technology.

TensorFlow-DirectML is a machine learning technology designed for a wide range of hardware including GPUs. As you can see, in the case of fresh AMD adapters, the performance can be very high. In general, the technology works on any graphics card with DirectX 12 support, and Microsoft itself mentions Nvidia and Intel as well.

Post a Comment