Breaking all previous AI performance records of six industry top MLPerf benchmarks, Google’s latest development is the world’s fastest machine learning (ML) training supercomputer.

By this supercomputer and the latest Tensor Processing Unit (TPU) chip, Google sets extraordinary performance records.

On Wednesday, Naveen Kumar Senior Staff Software Engineer / Sr. Engineering Manager said that they achieved the results with Machine Learning model implementations in TensorFlow, JAX, and Lingvo. Four from eight models got trained from root-level within just 30 seconds.

The results were achieved after three weeks of training of the models on the latest hardware accelerator available in 2015.

The latest TPU supercomputer trained the equivalent model almost five orders of magnitude faster within five years.

Models of MLPerf are selected as representative of cutting-edge machine learning workloads which are common in industry and academia.

MLPerf’s training round used supercomputer Google which is four times larger than “Cloud TPU v3 Pod” and created three records in the previous competition.

Nvidia, the graphics giant stated that it delivered the world’s fastest Artificial Intelligence training performance in commercially available chips. It is a feat which will help big enterprises to solve the complex challenges in AI, Data Science, and Scientific Computing.

Nvidia’s A100 GPUs and DGC SuperPOD systems were announced as the world’s fastest commercially available products for AI training, as per to MLPerf benchmarks.

A100 Tensor Core GPU demonstrated the fastest performance per accelerator on the existing eight MLPerf benchmarks. The company stated, “The real winners are the customers that apply this performance today for transforming respective businesses rapidly and cost-efficiently with AI.”

A100 is the first processor-based on the Nvidia Ampere architecture that hit the market more than any other Nvidia GPU.

The world’s leading cloud providers like Amazon Web Services (AWS), Baidu Cloud, Microsoft Azure and Tencent Cloud and other major server makers like Dell Technologies, Hewlett Packard Enterprise, Inspur and Supermicro are set to help Nvidia to meet the strong demands for A100.