Intel introduced at the Six Five Summit an Infrastructure Processing Unit (IPU), a relatively new category of technology for better data center efficiency and manageability.
The Intel IPU was built with input from hyperscale cloud partners and is being used by some of them in the form of FPGAs, but will also potentially be used by communication service providers to reduce overhead and free up CPUs for better performance.
“It is a little like having a hypervisor in hardware that provides the infrastructure for your customer workloads,” said Guido Appenzeller, the CTO for the data platforms group at Intel, in a short video released on Monday.
“The IPU is very strategic to our cloud strategy,” he added.
Intel said it already has FPGA-based IPU platforms deployed at multiple cloud service providers and is testing its first ASIC IPU. More FPGA-based IPUs dedicated ASICS will be rolled out, but Intel did not provide any timeline.
At its most basic level an IPU is a programmable network device. An IPU frees up CPU cores by shifting storage and network virtualization functions previously done in software on the CPU do the IPU. A cloud provider can accelerate and network virtualization and security, Intel said.
Analyst Patrick Moorhead of Moor Insights and Strategy said that Intel has ranked the IPU at the same level of importance as at CPU or XPU, “which makes this a huge announcement and shows how important this is. This is an indication of the new Intel.” Moor Insights and Strategy runs the Six Five Summit.
Moorhead said the Intel IPU falls in a similar tech category as the Nvidia DPU, which came with the acquisition of Mellanox, as well as AWS’s Nitro with its Annapurna acquisition. Azure has been using Intel SmartNICs to accelerate networking and storage since 2015. He said Intel’s SmartNIC is the market share leader in the category.
RELATED: Intel invests $3.5 B in N. Mexico chip expansion, atop Arizona plans