Google has revealed the details of a new version of its artificial intelligence chips for data centers and announced an Arm-based central processor.
An Arm processor is a type of CPU that uses RISC architecture, simplifying the instructions that the computer needs to process. Google’s Tensor Processing Units (TPUs) are one of the few alternatives to Nvidia’s advanced AI chips. However, developers can only use these via Google’s Cloud Platform and cannot purchase them directly.
However, Google’s new Axion CPU will initially support the company’s AI operations before becoming available to Google Cloud’s business customers later this year. The company said its performance is better than x86 chips and general-purpose ARM chips in the cloud. The Axion chips are used to serve YouTube ads, power the Google Earth Engine, and power various other Google services.
And @ThomasorTK announces #Axion Processor – @Googleis the first @POOR based chip – already sees 60% better energy efficiency than x86 #GoogleCloudNext pic.twitter.com/JpEdpA7oNd
— @GoogleCloud #GoogleCloudNext Holger Müller #AI (@holgermu) April 9, 2024
The Axion Arm-based CPU will deliver 30 percent improved performance over “general-purpose Arm chips” and outperform Intel’s current processors by 50 percent.
“We are making it easy for customers to move their existing workloads to Arm,” Mark Lohmeyer, vice president and general manager of computing and machine learning infrastructure at Google Cloud, told Reuters. “Axion is built on open foundations, but customers using Arm anywhere can easily adopt Axion without redesigning or rewriting their apps.”
Lohmeyer also explained in a blog that the tech giant is improving its TPU AI chips: “TPU v5p is a next-generation accelerator specifically designed for training some of the largest and most sophisticated generative AI models.” The Alphabet subsidiary announced that the new TPU v5p chip is designed to operate in 8,960 chip pods and delivers twice the raw performance of the previous TPU generation. The TPU v5p is already available via Google’s cloud.
Google’s new cloud AI hypercomputer architecture features
Google said it has significantly improved its hypercomputer architecture, focusing on performance-optimized hardware improvements. This includes the general availability of Cloud TPU v5p and A3 Mega VMs powered by NVIDIA H100 Tensor Core GPUs. These updates are intended to provide increased performance for large-scale training and improved networking capabilities.
The company has optimized its storage portfolio for AI workloads with the launch of Hyperdisk ML, a new block storage service designed for AI inference/deployment workloads. Additionally, new caching features were introduced in Cloud Storage FUSE and Parallelstore that improve training and inference on throughput and latency.
In the software sector, Google has introduced several open source developments. This includes JetStream, a throughput and memory optimized inference engine for large language models (LLMs) that delivers higher performance per dollar on open models like Gemma 7B.
Google is also introducing new flexible consumption options to better meet varying workload needs. This includes the Dynamic Workload Scheduler, which features a calendar mode for startup time backup and a Flex startup mode for optimized economics, further increasing the efficiency and flexibility of Google’s cloud computing offerings.
Featured Image: Canva