Google Cloud Expands Its Lineup of AMD-Powered Instances

Cloud

Google Cloud announced this week that it is expanding its lineup of virtual machine instances powered by AMD’s 3rd Gen EPYC processors, codenamed Milan. The new AMD-based VMs aim to provide increased performance and cost savings for a variety of compute-intensive workloads.

The new lineup includes general-purpose VMs for applications like web servers, backend services, interactive applications, gaming servers, and microservices. It also includes compute-optimized VMs for high-performance computing, batch processing, video encoding, distributed analytics, and ad serving.

Additionally, Google Cloud is launching memory-optimized VMs with large memory configurations for applications like high-performance databases, distributed memory caches, and in-memory analytics. The new lineup also includes AMD-based VMs optimized for storage performance and capacity.

According to Google Cloud, the Milan-based VMs can deliver up to 34% better price-performance over comparable Intel-based VMs for scale-out applications. For memory-intensive applications, the new AMD-powered memory-optimized VMs provide 46% more memory per vCPU than comparable Intel-based offerings.

Some key highlights of Google Cloud’s expanded AMD-based VM lineup:

– New C2D machine family with up to 96 vCPUs and 14.4 TB of memory
– New C2 machine family with up to 64 vCPUs and 9.6 TB of memory
– M2 machine family with large memory configurations up to 38.5 TB
– New E2 machine family optimized for storage performance

The C2D family offers high core counts in a dense configuration, making it well-suited for high-throughput scale-out applications and analytics. The large memory VMs in the M2 family provide an alternative to in-memory databases and caches. The E2 family is designed for data warehousing, distributed file systems, and other data-intensive workloads.

To complement the new instances, Google Cloud is introducing various AMD-powered Shielded VM options. Shielded VMs leverage advanced security features built into AMD EPYC processors. This helps provide protection against credential theft, rootkits, debug breakpoints, and more.

The expanded AMD-based lineup is part of Google Cloud’s Confidential Computing portfolio. Confidential Computing leverages hardware security capabilities to encrypt data in use. This helps mitigate unauthorized access or modification of data while it’s being processed.

Several Google Cloud products and services will leverage the new AMD-based VMs to provide benefits like better price-performance and more TDP (thermal design power) headroom for higher and more consistent CPU performance.

For example, Google Kubernetes Engine, BigQuery, and Cloud AI will utilize C2D and C2 machine types. Cloud Bigtable will take advantage of high memory M2 VMs. Dataproc will utilize general-purpose C2 VMs for scale-out applications like Apache Spark.

The announcement of expanded AMD-powered instances comes as Google Cloud continues trying to catch up to rivals Amazon Web Services and Microsoft Azure. Google Cloud currently has an estimated 7% market share in cloud infrastructure services spending, compared to 45% for AWS and 20% for Azure.

However, Google Cloud has been growing rapidly, with revenues up 37% in the first quarter of 2022. The addition of EPYC Milan-based instances provides Google Cloud with a powerful and cost-effective new lineup to continue attracting customers.

AMD launched its 3rd Gen EPYC processors in March 2021. The EPYC 7003-series processors utilize TSMC’s 7nm manufacturing process and feature up to 64 Zen 3 cores per chip.

Compared to the previous generation EPYC Rome processors, Milan offers up to 19% more instructions per cycle per core. This provides significant performance gains in both compute and memory-intensive workloads.

The advanced 7nm process allows EPYC Milan processors to deliver these performance gains while increasing core counts and boosting memory support and bandwidth. For virtualized workloads, EPYC Milan supports up to 400 GB per socket, 4TB per 4-socket server, and up to 64 cores and 128 lanes of PCIe 4.0.

Google Cloud first announced AMD-based VMs in 2020, starting with general-purpose N2D VMs powered by AMD’s 2nd Gen EPYC Rome processors. AWS and Azure also offer EPYC Rome instances. The new Milan-based VMs provide cloud providers with the latest server CPU advancements from AMD.

According to Forrester VP J.P. Gownder, “AMD’s processor innovation puts pressure on Intel to counter with its own advances.” This competition between AMD and Intel in the data center market has resulted in continuous innovation, providing cloud providers like Google Cloud more options for high-performance infrastructure.

The Google Cloud AMD-based VMs powered by EPYC Milan are available in all regions except for Germany. Google Cloud’s lineup of Intel-based VMs also continues to be available globally. The AMD and Intel instances provide varied options across price, performance, core count, and memory configurations.

Looking ahead, Google Cloud customers can expect more innovation around Confidential Computing, Shielded VMs, and security capabilities enabled by AMD EPYC processors. Google Cloud will likely continue expanding its AMD-powered VM lineup as AMD releases next-gen EPYC Genoa processors in 2022 on 5nm technology.

Overall, the new Milan-based instances demonstrate Google Cloud’s commitment to providing a breadth of VM options optimized for different customer needs. With significant price-performance benefits over comparable Intel VMs, the new AMD-powered lineup gives Google Cloud fresh capabilities to meet demanding scale-out computing, HPC, analytics, database, and other enterprise workloads.

Leave a Reply

Your email address will not be published. Required fields are marked *