At Google, we regularly push the limits of computers to investigate what can be done for big problems like global video distribution, information retrieval, and, of course, generative AI. Rethinking systems design in close cooperation with service developers is necessary to achieve this. Our large investment in bespoke silicon is the outcome of this rethinking.
We are excited to present the most recent iteration of this effort today: Google Axion Processors, our first specially made Arm-based CPUs intended for data centers. Later this year, Axion—which offers performance and energy efficiency that leads the industry—will be made available to Google Cloud users.
Axion is only the most recent model of customized silicon from Google. In 2018, we introduced our first Video Coding Unit (VCU), achieving up to 33x more efficiency for video transcoding. Since 2015, we have released five generations of Tensor Processing Units (TPUs). In 2021, we doubled down on custom computing by investing in “system on a chip” (SoC) designs and releasing the first of three generations of Tensor chips for mobile devices.
General-purpose computing is and will continue to be a vital component of our customers’ workloads, even if our investments in compute accelerators have revolutionized their capabilities. Extensive computation power is needed for analytics, information retrieval, and machine learning training and providing.
The rate at which CPUs are being improved has slowed recently, which has affected customers and users who want to satisfy sustainability goals, minimize infrastructure costs, and maximize performance. According to Amdahl’s Law, unless we make the corresponding expenditures to stay up, general-purpose computing will dominate the cost and limit the capabilities of our infrastructure as accelerators continue to advance.
To deliver instances with up to 30% better performance than the fastest general-purpose Arm-based instances currently available in the cloud, as well as up to 50% better performance and up to 60% better energy efficiency than comparable current-generation x86-based instances, Axion processors combine Google’s silicon expertise with Arm’s highest performing CPU cores.
For this reason, on current generation Arm-based servers, we have already begun implementing Google services such as BigTable, Spanner, BigQuery, Blobstore, Pub/Sub, Google Earth Engine, and the YouTube Ads platform. We also have plans to deploy and expand these services, along with others, on Axion shortly.
High effectiveness and performance, supported by titanium
Axion processors, which are constructed around the Arm NeoverseTM V2 CPU, offer massive performance gains for a variety of general-purpose workloads, including media processing, web and app servers, containerized microservices, open-source databases, in-memory caches, data analytics engines, and more.
Titanium, a system of specially designed silicon microcontrollers and tiered scale-out offloads, provides the foundation for Axion. Platform functions like networking and security are handled by titanium offloads, giving Axion processors greater capacity and enhanced performance for workloads from customers. The titanium also transfers I/O processing for storage to Hyperdisk, our recently launched block storage solution that can be dynamically supplied in real-time and decouples performance from instance size.
“Building on our high-performance Arm Neoverse V2 platform, Google’s announcement of the new Axion CPU represents a significant milestone in the delivery of custom silicon optimized for Google’s infrastructure.” Decades of ecosystem investment, together with Google’s continuous innovation and contributions to open-source software, provide the optimal experience for customers operating on Arm workloads that are most important.
Customers want to accomplish their sustainability goals and operate more effectively, not only perform better. In comparison to five years ago, Google Cloud data centers are now 1.5 times more efficient than the industry average and provide 3 times more processing power with the same amount of electrical power2. Our lofty objectives include running our campuses, offices, and data centers entirely on carbon-free energy sources around the clock and providing resources to assist with carbon emission reporting. Customers may optimize for even greater energy efficiency with Axion processors.
Axion: interoperability and compatibility with out-of-the-box applications
Additionally, Google has a long history of supporting the Arm ecosystem. We collaborated closely with Arm and industry partners to optimize Android, Kubernetes, TensorFlow, and the Go language for the Arm architecture. We also constructed and made them open-sourced.
The standard Armv9 architecture and instruction set serve as the foundation for Axion. We have made contributions to the SystemReady Virtual Environment (VE) standard, which is designed to ensure that Arm-based servers and virtual machines (VMs) can run common operating systems and software packages. This standard makes it easier for customers to deploy Arm workloads on Google Cloud with minimal to no code rewrites. We are gaining access to an ecosystem of tens of thousands of cloud users who are already using Arm-native software from hundreds of ISVs and open-source projects and deploying workloads thanks to our partnership.
Axion will be available to users across a variety of Google Cloud services, such as Cloud Batch, Dataproc, Dataflow, Google Compute Engine, and Google Kubernetes Engine. The Google Cloud Marketplace now offers Arm-compatible apps and solutions, and we just released preview support for the Migrate to Virtual Machines service, which allows you to migrate Arm-based instances.
What our partners and consumers have to say
“We’re excited to introduce application packages to the Bitnami by VMware Tanzu library that are designed for Google Cloud that utilize the new Axion Arm-based CPU. For our users, this will result in noticeably higher performance, an appealing price-performance ratio, and more sustainability. We can’t wait to use the new Google Axion instances and go above and above to support our clients in streamlining installations and lessening their environmental impact.”
“To prevent cloud breaches, businesses worldwide rely on CrowdStrike and its one platform, single agent design. As CrowdStrike has the shortest deployment times in the market and the strongest security, we’re eager to test Google’s new CPU to see if there are any power and efficiency advantages.” — Daniel Bernard, CRO of Strike’s Chief Business Officer
“Our clients expect our systems to offer unwavering cybersecurity protection. We’re excited about the potential power and efficiency benefits offered by the new Google Cloud’s unique Arm-based CPU, and we want to verify its capabilities to give our customers even greater threat detection and response capabilities. VP of Business Development at Cybereason, Tzach Segal
“Datadog was an early adopter of Arm for our operations and a dependable partner for clients using Arm-based virtual machines. We’re thrilled that Google Cloud has announced the Axion processor, and we intend to test it on our workloads while assisting clients in calculating the advantages of running Datadog on Google Cloud Arm instances in terms of both cost and performance. Yrieix Garnier, datadog’s vice president of product management
“Elastic is dedicated to assisting clients in effectively utilizing all of their organized and unstructured data at any size. We anticipate using Google Cloud’s new proprietary Arm-based CPU to deliver an even better Elastic customer experience on the Google Cloud Platform, and we are eager to test it. -Elastic Vice President of Product Steve Kearns
Over the years, we’ve developed a solid working relationship with Google Cloud and have witnessed the advantages of utilizing GCP Arm-based virtual machines. We eagerly anticipate witnessing the amazing advancements that the next generation of Arm-based Axion CPUs will bring. – Joel Meyer, OpenX’s SVP of Engineering
“With Snap, everyone can express themselves, be present, discover the world, and have fun with one another. Our infrastructure is always being improved for efficiency and performance.
Significant advancements in both are promised by Google’s new Axion Arm-based CPU. It is very exciting to have the opportunity to lead on our sustainability goals and use these advantages to benefit our community. When Axion-based virtual machines are released, we’re eager to see their advantages.” – Senior Director of Engineering, Cloud Infrastructure, Korwin Smith, Snap.
In 150 countries, the WordPress Engine powers websites for over 1.5 million users. They count on WP Engine to fulfill our three main commitments of security, dependability, and performance, and we take that duty very seriously. Because of our dedication to innovation and customer-inspired thinking, we’re constantly searching for new and improved methods to help our clients succeed and feel more confident online. We can’t wait to test the predicted performance and sustainability improvements of Google’s new bespoke Arm-based processor and see how they will enable our clients to build even more engaging web pages and applications. – Ramadass Prabhakar, WP Engine’s SVP and CTO.
