Google launches VMs powered by newest Intel Xeon

Google Cloud announced new virtual machines as part of its cloud platform based on Intel’s newest Xeon Scalable processors.

In addition to the new C3 virtual machine series, Google also announced it is deploying Infrastructure Processing Units (IPU), which are designed to intelligently route network traffic and take the load of network data processing off the CPU. The  company made the announcements at its Google Cloud Next ’22 conference held virtually.

The IPU chip, formally known as the E2000, was co-designed by Google and Intel together and features 16 Arm Neoverse cores and 200GbE networking.

C3 machine instances deliver performance gains of up to 20% over previous generation of C2 instances. They also benefit from a recent product launch called Hyperdisk, a block storage system that offers 80% higher IOPS per vCPU for data analytics and DBMS workloads when compared to other hyperscalers,

All told, Google says C3 VMs with Hyperdisk deliver four times higher throughput and 10 times greater IOPS when compared with the previous C2 generation. It makes the C3 instances capable of offering high-performance computing (HPC) levels of performance, something cloud-based VMs are not known for.

Intel launched the IPU last year. The concept is basically the same as what its silicon competitors (AMD, Nvidia, Marvell) call data processing units (DPUs), namely lightening the burden on CPUs. IPUs are designed to offload functions like routing network traffic, handling traffic analysis and storage and network virtualization.

The new Xeon CPUs (codenamed Sapphire Rapids) have been the subject of considerable delay due to manufacturing problems. They were originally slated to launch in 2021, but Intel is still struggling to bring them to market.

Gooogle Cloud’s C3 family of VMs are now available in private preview. General availability has not been announced.