Google Cloud Boosts Arm Efficiency with New Axion N4A Chip

Google Cloud Expands Arm-Based Computing with Axion N4A

With the release of Axion N4A, Google Cloud is stepping up its use of Arm-based computing in an effort to draw in businesses trying to cut costs and lessen reliance on x86-based instances.

Its N series of compute instances are usually intended for general-purpose workloads like small-to-medium databases, containerized microservices, virtual desktops, back-office, CRM, or BI applications, data pipelines, and web and application servers with low to medium traffic.

Currently in preview, the N4A instances allow up to 64 vCPUs, 512GB of DDR5 memory, and a 50 Gbps networking connection.

Axion N4A Promises Better Performance and Affordability

Google claims that N4A is the most affordable N series instance to date, with “up to 2x better value for money and 80% better performance per watt compared to comparable current-generation x86-based VMs.”

According to Brandon Hoff, research director at IDC, the company is able to offer such cost savings because it developed the Arm-based Axion processors internally and sources them directly from manufacturers like Taiwan Semiconductor Manufacturing Company (TSMC), thereby halving the cost of the underlying hardware. This, along with the chips’ performance, makes Google’s Arm-based instances an attractive option for businesses, he said.

Read: Instacart Launches AI Cart Assistant, Sprouts Leads Adoption

Growing Adoption of ARM Chips Across Major Cloud Providers

Google already uses Arm instances internally for applications like YouTube, Gmail, and BigQuery because the savings are so alluring.

It’s not alone: 50% of new AWS instances now use Graviton, an ARM-based chip that AWS released in 2018 to lower the cost of operating internal cloud workloads like Amazon retail IT. In order to run Microsoft 365 and provide Azure services, Microsoft also recently created the Cobalt Arm processor.

Availability and Future Plans for Google’s N4A Instances

Google’s N4A instances will be accessible through a variety of services, such as Dataproc for big data and analytics, Google Kubernetes Engine (GKE) for containerized workloads, and Compute Engine for direct virtual machine execution. Initially, they will be available in the US-central1 (Iowa), US-east4 (N. Virginia), Europe-west3 (Frankfurt), and Europe-west4 (Netherlands) regions.

N4A, according to the business, is an addition to the C4A instances it introduced in October of last year. These are made for more demanding workloads like CPU-based AI and machine learning, high-traffic web and application server environments, ad servers, game servers, data analytics, and databases of any size.

C4A Metal, a bare-metal instance for specialised workloads in a non-virtualized environment, including custom hypervisors, security workloads, or CI/CD pipelines, is also “soon” to be released.