Fudzilla had a chance to chat with Lee Rossiter, who is in charge of the professional deep learning market and its customers. As this market started only three years ago, this is the right time to jump as the research is led by key universities, and important companies including Microsoft, Google, Facebook and Baidu see great potential in the deep learning market.
Nvidia managed to get even very conservative companies such as SAP excited about deep learning and because you will be able to recognize and use optimized images and speech, you will be able to use the same technology to improve processes in a company.
Scan.co.uk realized that it is a good idea to jump into this market as the educational market, startups, oil and gas industry as well as the financial markets are looking into using deep learning computers.
Scan will offer a bundle for early adopters and the one that caught our attention was the ability to rent a DXG-1 System with eight P100 powered by Pascal chips and 16GB HBM 2 memory, summing up to a total horsepower of 28672 Cuda cores or 170 teraFLOPS of half-precision (FP16) peak performance.
Nvidia DGX-1 is likely to be one of the fastest deep learning supercomputers on the market and they will sell for more than $100,000 even to top educational clients. The rest will probably end up paying even more.
Deep learning uses the parallel process abilities of Nvidia GPUs to reproduce the functions of a brain in a deep neural network in order to detect patterns and concepts in data. This process, due to its incredible ability to parallelize, can dramatically speed up speech recognition, language translation, image recognition and trend analysis.
There are also proof of concept option as well as the DGX-1 rental option.