Rack Mount Deep Learning Development Machine

■ This is an article posted on June 2016, 8, so the content of the information may be out of date.

We received an inquiry from a customer who is considering a deep learning development machine for joint research. The GTX 980 GPU has been specified according to the customer's current development environment. In addition, we are requesting to use Caffe (Deep Learning framework), CUDA, and mount it on the 19-inch server rack we are currently using.

As for the case, we have specified a GPGPU for desktop workstations, so we proposed a rack-mountable 4U tower case due to the mounting structure and cooling. If you choose a server board regardless of GTX 980, we can also provide 1U to 2U.

In addition, we have also specified the OS and detailed settings at the time of shipment, so this is a guide assuming each of them.

【Main Specifications】

CPU Xeon E5-1650v4 (3.60GHz 6 cores)
memory 64GB ECC REG
storage 480GB SSD/1TB HDD
network Gigabit LAN x2
video GTX 980
Housing + power supply 4U tower enclosure (using OP rail) + 850W
OS Ubuntu 14.04LTS Desktop 64bit (English version / No Japanese environment)
Others CUDA Toolkit 7.5 Install