After unveiling the Tesla P100 HPC accelerator, Nvidia also unveiled the DGX-1 Deep Learning System which actually comes with eight Tesla P100 HPC accelerator, claiming the "world's first deep-learning supercomputer in a box."

Based on eight Tesla P100 HPC accelerator connected via Nvidia's NVLink interconnect, the DGX-1 supercomputer is, according to Nvidia, up to 75 times faster than a dual-socket Xeon E5-2697 v3 system when running deep-learning tasks.

The Nvidia DGX-1 Deep Learning System provides up to 170 TFLOPS of compute performance and is up to 12 times faster in neural-training tasks compared to a server based on four Maxwell GPUs.

The rest of the specifications for the Nvidia DGX-1 Deep Learning System include two 16-core Xeon E5-2698 v3 CPUs, 512GB of DDR4 RAM, four 1.92TB SSDs in RAID-0 array, dual 10Gb Ethernet ports and a power draw of 3200W.

All these specifications also come at a rather high price so the DGX-1 will be available in June for US $129,000.







Source: Nvidia.com.