Microsoft Adds Intel FPGAs to Azure Cloud for AI
Intel and Microsoft are again joining forces, this time to accelerate Azure Cloud machine learning tools, including AI inferencing models, based on Intel’s FPGAs.
Microsoft also this week announced the integration of its Project Brainwave, a hardware architecture designed to accelerate real-time AI calculations, with its Azure machine learning platform.
Microsoft unveiled Azure machine learning models during a company event on Monday (May 7), adding that a software development kit available for preview could be accessed via its large cloud deployment of Intel FPGAs. The combination would allow data scientists and developers to train machine learning models, then deploy them on Microsoft’s Project Brainwave.
Intel FPGAs along with Xeon processors could then be used either in the cloud or at the network edge, they added.
The development kit for the Python programming language is designed to allow developers to focus on image-recognition tasks achieved by using their data to retrain deep neural networks based on ResNet-50models. (The industry standard deep neural network requires nearly 8 billion calculations.) FPGAs running on the Azure cloud would allow developers to tune hardware for precise retraining tasks associated with real-time AI workloads, the partners said.
Microsoft (NASDAQ: MSFT) said a limited preview of Project Brainwave provides access to FPGA-based systems running on-premises. The configuration serves as an Internet of Things edge device that connect to an IoT hub running on the Azure cloud.
“These new capabilities will allow the integration of AI into real-time processes,” said Doug Burger, a Microsoft distinguished engineer.
The partners also are looking to provide potential customers with greater flexibility in running cloud workloads. “One big benefit that FPGAs have is that they can be hardware programmable,” noted industry analyst Patrick Moorhead. “One day they can be doing deep packet network inspection and the next hour they can be doing machine learning for inferencing.”
Microsoft noted that Project Brainwave represents the leading edge of an effort to make cloud-based FPGAs available for a range of applications. “I think this is a first step in making the FPGAs more of a general-purpose platform for customers,” Mark Russinovich, CTO of Microsoft Azure cloud computing platform, noted in a blog post.
Meanwhile, the collaboration with Microsoft is Intel’s latest FPGA push. In April, the chip maker (NASDAQ: INTC) announced adoption of a programmable acceleration card that combines its Arria 10 GX FPGA and Xeon CPU within the server lineups of Dell-EMC and Fujitsu.