The custom silicon, dubbed the Azure Boost DPU, was unveiled at Microsoft’s Ignite event. It’s designed to run intense data workloads with improved efficiency and low power by running a lightweight data-flow operating system to enhance performance and lower power consumption.
Microsoft said it expects the new DPUs to run cloud storage workloads using three times less power and four times the performance compared to existing CPUs.
Subscribe today for free
“Azure Boost DPU integrates high-speed Ethernet and PCIe interfaces along with network and storage engines, data accelerators, and security features, into a fully programmable system on chip,” a Microsoft blog post authored by Pradeep Sindu, CVP for silicon at Microsoft, reads.
Microsoft has been working on the networking chip project for some time, adding to its growing line of custom silicon, which includes CPUs Maia and Cobalt.
Its new Azure Boost DPUs can optimise networks, enabling performance to remain efficient during intense workloads such as AI training.
Microsoft’s Azure Boost DPUs adds to the growing DPU landscape, with AMD also unveiling two new offerings, the Pensando Salina 400 and Pensando Pollara 400, back in October.
In addition to the custom DPUs, Microsoft also unveiled the Azure Integrated Hardware Security Module (HSM), an in-house developed security chip.
The custom hardware will be installed in every new server in Microsoft’s data centres, adding an extra layer of protection by allowing encryption and signing keys to remain on the hardware — without impacting performance or latency.
The security chips will begin installation in 2025.
“Azure Integrated HSM eliminates the classic tradeoff between increased network round trip latency to remote HSM services or seeking the release of keys from the remote HSMs,” Mark Russinovich, CTO, deputy CISO and technical fellow at Microsoft Azure, wrote in a blog post.
“As a server-local HSM that securely binds to the workload environments, Azure Integrated HSM provides locally attached HSM services to both confidential and general-purpose virtual machines and containers,” Russinovich wrote. “This provides the benefit of industry-leading in-use key protection without the latency drawbacks of round-trip network-attached HSM calls.”
RELATED STORIES
Microsoft’s got wood: Hyperscaler constructs timber-based data centres
Amazon launches $110m project to provide custom chips for AI research
