HPC & Datacenter

Changing Compute Requires New Compute Elements

As machine learning is rapidly becoming a common part of data center operation, new types of devices are needed to offload the main compute resources. We’re seeing the emergence of data-processing unit (DPUs), smart NICs (network interface cards) and NAS (Network Attached Storage) to handle this operation, as well as more and more AI accelerators within compute elements themselves, and in the storage and networking segments of the data center.

MIPS provides scalable multiprocessors designed to support the transition of the data center to new types of computation such as machine learning acceleration. MIPS solutions offer scalable performance and heterogenous capabilities that enable each cluster of compute resources to include unique combinations of CPUs and AI accelerators. We also provide optimized data-movement solutions for efficient handling of large amounts of data.