Admin

Google creates ‘Axion’ CPU for datacenters internally

ARM, Axion, CPU, Datacenters, Develops, Google, In-House



Google has long been known for its advancements in technology, particularly in the field of artificial intelligence (AI). The company was one of the first to develop custom silicon chips for its services, such as tensor processing units (TPUs) for AI initiatives and video transcoding units (VCUs) for YouTube. However, when it came to custom CPU designs, Google had been slower to adopt them, instead choosing to use off-the-shelf chips from major CPU manufacturers. This is finally changing with the introduction of Google’s new in-house datacenter CPU, the Axion.

The Axion processor is based on the Arm Neoverse V2 (Arm v9) platform, which is Arm’s current-generation design for high-performance server CPUs. This platform is already used in other chips, such as NVIDIA’s Grace. Within Google, the Axion CPU is designed for a wide range of workloads, including web and app servers, data analytics, microservices, and AI training. Google claims that the Axion processors offer up to 50% higher performance and up to 60% better energy efficiency compared to current-generation x86-based processors. They also claim a 30% higher performance compared to other Arm-based CPUs for datacenters. However, Google does not specify which specific processors they are comparing Axion to in these metrics.

While Google has not disclosed the core counts or full specifications of the Axion CPUs, they have revealed that the chips incorporate their own “secret sauce” in the form of Titanium purpose-built microcontrollers. These microcontrollers are designed to handle basic operations such as networking and security, as well as offload storage I/O processing to the Hyperdisk block storage service. By offloading these tasks, the CPU core resources are freed up for actual workloads. In terms of memory, Axion uses conventional dual-rank DDR5 memory modules.

The development of the Axion CPU marks an important milestone for Google in delivering custom silicon optimized for their infrastructure. According to Rene Haas, CEO of Arm, Google’s investment in the Arm ecosystem, along with their ongoing innovation and open-source contributions, ensures the best experience for customers running on Arm architectures. This implies that Google’s choice to develop its own CPU is not only driven by the desire for better performance and efficiency but also by the ability to fully optimize the silicon for their specific workloads.

Google has previously used Arm-based processors for some of its own services, including BigTable, Spanner, BigQuery, and YouTube Ads. With the introduction of the Axion CPU, Google is now ready to offer instances based on Armv9 architecture to its customers. These instances will be able to utilize software developed for Arm architectures, opening up new possibilities for developers and users.

In conclusion, Google’s development of the Axion CPU represents a significant step forward in their quest for custom silicon optimized for their datacenters. By utilizing the Arm Neoverse V2 platform and incorporating their own purpose-built microcontrollers, Google aims to deliver higher performance and energy efficiency compared to traditional CPUs. With the ability to offload basic operations and optimize core resources, the Axion CPU has the potential to greatly enhance the capabilities of Google’s infrastructure. This move also demonstrates Google’s commitment to Arm-based architectures and their efforts to contribute to the Arm ecosystem. Overall, the Axion CPU is an exciting development that showcases Google’s dedication to pushing the boundaries of technological innovation.



Source link

Leave a Comment