NVIDIA DIGITS DEVBOX Promotes Deep Learning w/ Titan X


NVIDIA announced the $15,000 Deep Learning GPU Training System (DIGITS) at GTC 2015 to become a NVIDIA’s very first deep learning development box for data scientists and researchers. NVIDIA saw a need for a powerful system that could used by professionals to build a deep neural network (DNN) for whatever real-time network data set you are wanting to use. To quickly create high-quality deep neural network you need something that can handle the highly parallel workloads, so  NVIDIA went with four NVIDIA GeForce GTX Titan X graphics cards (our review) to get the most performance possible for their customers. A Quad-SLI setup is the most one can do with the Titan X, so this is a very powerful single PC. DIGITS uses open-source software, available on GitHub on Ubuntu 14.04, so developers can extend or customize it or contribute to the project. The only supported Linux distribution right now is Ubuntu 14.04.

NVIDIA Digits Demo System at GTC 2015

The NVIDIA DIGITS DEVBOX is built with mostly off the shelf components and we were able to see a couple of the systems in the exhibit hall at GTC 2015.

nvidia digits devbox titan x

From what we gather NVIDIA is building these systems with high-end off the shelf parts. We found the system on display featured an ASUS X99E-WS motherboard with what we expect would be an Intel Core i7 Haswell-E processor. Other goodies inside the Corsair Carbide Series Air 540 chassis included 64GB of Corsair DDR4 memory, a 120mm water cooler for the processor, a three 3TB RAID 5 array for the storage disks and finally a massive 1500 Watt power supply.


The NVIDIA GeForce GTX Titan X 12GB graphics cards run $999 each and are only available for purchase from the GeForce shop right now.


One of the NVIDIA DIGITS was hooked up to the demo and the other one demo system with the case door off was not networked and was just for show.


The NVIDIA DIGITS might look like the ultimate gaming PC, but it was designed for researchers to be able to work on deep learning to do real meaningful work with very little setup time. In fact, NVIDIA says that you should be able to start using a DIGITS DEVBOX in under an hour!