Nvidia has shown off the world’s largest GPU at its GTC conference.
The mega GPU was unveiled during the two-hour keynote speech by Nvidia CEO Jensen Huang, along with a number of other upcoming products including a cloud supercomputer named Clara. The Nvidia DGX-2, a GPU supercomputer, will become Nvidia’s most expensive piece of kit with pricing fetching as much a $400,000. Weighing in at a hefty 350 pounds, the DGX-2 will feature 16 individual Tesla V100 GPUs, each with 32GB of system memory. These cards all work together through Nvidia’s NVSwitch technology, and our designed for deep-learning applications which require massive amounts of processing power.
Huang said that the new supercomputer had been created for researchers who are ‘pushing the outer limits of deep-learning research and computing’ to train artificial intelligence."Deep neural networks are getting deeper and smarter and able to do more things," Huang said in a press briefing. "As they grow in size and complexity, so do the computing needs to train them."
Nvidia also unveiled its Clara cloud supercomputer at the conference, briefly demonstrating Clara’s capabilities. The system is designed to enable doctors and medical professionals working with old or outdated equipment to provide better care to patients without having to invest in multi-million-dollar imaging technology. The system, as demoed, can take the standard data output from a 15-year-old ultrasound machine, run it through its artificial intelligence models and actually infer more information than the ultrasound originally contained. During the demo, Clara took a simple two-dimensional image of a heart beating and turned it into a fully three-dimensional image, complete with blood pressure estimates and other health diagnostics.