Jen-Hsun Huang, CEO for the California based tech company, Nvidia Corporation, has unveiled on Tuesday a new GPU, entitling DGX-1. The unveiling has been made during Nvidia's GPU Technology Conference in San Jose, California.
DGX-1 is basically a fancy enclosure for an 8-GPU supercomputing cluster. It contains 8 Tesla P100 cards, each with 16 GB of RAM and a 7 TB memory for raw data storage to be used by deep learning networks.
There is also a NVLink Hybrid Cube Mesh and in-built neural network training software. Interested researchers and companies may be able to roll with their own solutions replacing the software.
A standardized platform like DGX-1 may perform guesswork out of a building system. Furthermore, NVidia's support and regular update justify the extra cost for the budget conscious consumers, reports TechCrunch. Chips inside the DGX-1 are arranged in a hybrid cube array, allowing them to communicate with each other independently and carry out commands in a coordinated manner. It offers a too simple solution for developing an AI network in any enterprise and is capable of performing a task thrown at it in shortest order, according to a report published in Digital Trends. The Tesla P100 cores inside the supercomputer are capable of delivering 170 teraflops of half-precision (FP16) peak performance which is equivalent to 250 CPU based servers. The supercomputer will cost $129,000. The tech giant has committed to ship the initial phase produce to universities and hospitals engaged in AI research including Massachusetts General Hospital (MGH), reports Pc Magazine. The DGX-1 is already available for pre-order and shipping will start during the third quarter. However, shipments of the hospital and university versions are expected to roll out any time soon. MGH is going to enter the radiological era of biometric qualification with enhanced interpretations through algorithms learned from the diagnostic data of vast patient populations, cites Dr. Keith J. Dreyer, an MGH radiologist through a statement. He, however, acknowledges the processing capabilities of the GPUs as the key factor behind the dreamed revolution in radiological science. The Californian tech giant is completely committed to artificial intelligence and deep learning which is getting evident from unveiling more intense high-end products. Analysts presume that the GPUs of tomorrow will be built for AI and deep learning while gaming will be considered just as a hobby.
Join the Conversation