NVIDIA DGX-1, World’s First Deep Learning Supercomputer Launched
Calling it the first supercomputer in a box, Nvidia has launched what they are also calling the first deep learning supercomputer in the form of the NVIDIA DGX-1. While it does not look like it from the design, it is one beast of a machine.
Boasting of 170 teraflops of peak performance at half-precision (FP16), this machine is dedicated and specially crafted for artificial intelligence, deep learning and other research aspects that cannot be done without a massive load of computing power.
NVIDIA DGX-1 is The World’s First Deep Learning Supercomputer
With eight Tesla GP100 GPUs of 16GB each, 512 GB of RAM and four 1.92TB of SSD, this thing weighs 60kgs which is equivalent to 132 lb and uses Nvidia Pascal architecture.
The NVIDIA DGX-1 deep learning system is built on NVIDIA Tesla® P100 GPUs, based on the new NVIDIA Pascal™ GPU architecture. It provides the throughput of 250 CPU-based servers, networking, cables and racks — all in a single box.
The DGX-1 features four other breakthrough technologies that maximize performance and ease of use. These include the NVIDIA NVLink™ high-speed interconnect for maximum application scalability; 16nm FinFET fabrication technology for unprecedented energy efficiency; Chip on Wafer on Substrate with HBM2 for big data workloads; and new half-precision instructions to deliver more than 21 teraflops of peak performance for deep learning.
Here are the system specifications:
- Up to 170 teraflops of half-precision (FP16) peak performance
- Eight Tesla P100 GPU accelerators, 16GB memory per GPU
- NVLink Hybrid Cube Mesh
- 7TB SSD DL Cache
- Dual 10GbE, Quad InfiniBand 100Gb networking
- 3U – 3200W
As far as the availability is concerned, the Nvidia DGX-1 is going to be available in the US by JUne this year although the rest of the regions will have to wait until the third quarter.