To solve the world’s most profound challenges, you need powerful and accessible machine learning (ML) tools that are designed to work across a broad spectrum of hardware. This can range from datacenter applications for scientists and researchers to desktop and notebook PCs used by students and professionals looking to develop ML models on the hardware they already own.
The feedback from our customers is their desire to better leverage the power of their existing Windows® PC hardware for their machine learning workflows. To help address this need and make ML tools more accessible to Windows users, last year Microsoft announced the preview availability of support for GPU-accelerated training workflows using DirectML-enabled machine learning frameworks in Windows and the Windows Subsystem for Linux (WSL).
This GPU-accelerated training works on any DirectX® 12 compatible GPU and AMD Radeon™ and Radeon PRO graphics cards are fully supported. This provides our customers with even greater capability to develop ML models using their devices with AMD Radeon graphics and Microsoft® Windows 10.
TensorFlow-DirectML Now Available
To complement the support above, Microsoft released the TensorFlow-DirectML preview for Windows and the WSL, the TensorFlow software library being one of the most widely used tools for machine learning training workflows. Over the past year, Microsoft has continued to develop it, including open-sourcing the project, all leading to today’s release of the first generally consumable package of TensorFlow-DirectML.
Throughout the development of TensorFlow-DirectML, AMD and Microsoft have been working closely together. We have collaborated on co-engineering to deliver multiple improvements and performance optimizations that improve the GPU-accelerated ML training workflow experience when using TensorFlow-DirectML on DirectX 12 compatible AMD graphics hardware. In addition to these efforts, AMD has also continued to improve ML TensorFlow inference performance on select AMD Radeon GPUs.
Performance Improvements Across the Board
The performance optimizations have improved both machine learning training and inference performance. Using the AI Benchmark Alpha benchmark, we have tested the first production release of TensorFlow-DirectML with significant performance gains observed across a number of key categories, such as up to 4.4x faster in the device training score (1). And at AMD we will continue to seek out opportunities to optimize and performance tune AMD Radeon GPUs and continue to deliver great results for our customers.
For the AMD-specific ML performance improvements, we’ve updated our driver to deliver substantially better TensorFlow inference performance on AMD Radeon™ RX 6900 XT and RX 6600 XT graphics hardware. When tested with AI Benchmark Alpha and the release version of TensorFlow-DirectML, we saw up to a 3.1x increase in inference performance (2) with this update.
When you combine the work on both ML training and inference performance optimizations that AMD and Microsoft have done for TensorFlow-DirectML since the preview release, the results are astounding, with up to a 3.7x improvement (3) in the overall AI Benchmark Alpha score!
Start Working with TensorFlow-DirectML on AMD Graphics Today
The good news for those of you looking to have a great experience using TensorFlow-DirectML with DirectX 12 compatible AMD Radeon graphics hardware is that all the performance optimizations are already available in the latest AMD Radeon™ Software Adrenalin drivers which can be downloaded here.
To learn more and try out today’s production release of TensorFlow-DirectML and use your current Windows PC with AMD graphics hardware for TensorFlow machine learning training workflows, you can check out Microsoft’s blog for more information.
These updates will let you benefit from all the performance improvements that both Microsoft and AMD have enabled throughout the development of TensorFlow-DirectML and will take your Windows machine learning workflows when using AMD graphics to the next level!
Alexander Blake-Davies is a Software Product Marketing Specialist for Radeon Software for Developers at AMD. Links to third-party sites are provided for convenience and unless explicitly stated, AMD is not responsible for the contents of such linked sites and no endorsement is implied. GD-5
©2021 Advanced Micro Devices, Inc. AMD, the AMD Arrow logo, Radeon, Ryzen, and combinations thereof are trademarks of Advanced Micro Devices, Inc. Microsoft, DirectX, and Windows are registered trademarks of Microsoft Corporation in the United States and/or other jurisdictions. TensorFlow, the TensorFlow logo, and any related marks are trademarks of Google Inc. Other product names used in this publication are for identification purposes only and may be trademarks of their respective companies.