Nvidia delivers Tesla K80 GPU accelerator for analytics and supercomputing

Speeds up high-performance computing via 480GBps memory bandwidth

Nvidia has added a fresh GPU accelerator to its Tesla Accelerated Computing Platform, which the firm claims is the highest-performing of its class.

Designed for high-performance computing (HPC) applications and supercomputers , the Tesla K80 dual GPU accelerator is the most powerful in Nvidia's line-up and is aimed at accelerating a wide range of data analytics and scientific computing machine learning.

"[The Tesla K80] combines the world's fastest GPU accelerators, the widely used CUDA parallel computing model, and a comprehensive ecosystem of software developers, software vendors, and datacentre system OEMs," Nvidia said, adding that it delivers almost double the performance and double the memory bandwidth of its predecessor, the Tesla K40 GPU accelerator.

"With 10 times higher performance than today's fastest CPU, it outperforms CPUs and competing accelerators on hundreds of complex analytics and large, computationally intensive scientific computing applications," the firm added.

The accelerator boasts an enhanced version of Nvidia GPU Boost technology, which is said to boost applications by converting power headroom into the optimal performance enhancement for each individual application.

The GPU was designed to tackle "the most difficult computational challenges", which range from astrophysics and genomics to quantum chemistry and data analytics. It has also been made to take on deep learning tasks, a segment of the machine learning field that Nvidia said is the fastest growing.

Featuring two GPUs per board, the Tesla K80 dual-GPU accelerator doubles throughput of applications designed to take advantage of multiple GPUs. It offers 24GB of ultra-fast GDDR5 memory - 12GB of memory per GPU - as well as twice the memory of the Tesla K40 GPU, which allows the processing of double-sized datasets.

A 480GBps memory bandwidth means that data scientists can crunch through petabytes of information in half the time compared with the Tesla K10 accelerator.

Nvidia said that the GPU's 4,992 CUDA parallel processing cores also boost applications by up to 10 times compared with using a CPU alone.

Wolfgang Nagel, director of the Centre for Information Services and HPC at Technische Universität Dresden in Germany, said, "The Tesla K80 dual-GPU accelerators are up to 10 times faster than CPUs when enabling scientific breakthroughs in some of our key applications, and provide a low energy footprint.

"Our researchers use the available GPU resources on the Taurus supercomputer extensively to enable a more refined cancer therapy, understand cells by watching them live, and study asteroids as part of ESA's Rosetta mission."

The Tesla K80 dual-GPU accelerator is shipping now from server manufacturers including Asus, Bull, Cirrascale, Cray, Dell, Gigabyte, HP, Inspur, Penguin, Quanta, Sugon, Supermicro and Tyan.