Feedback!

Build a Deep Learning Linux Workstation

Views: 3198 Difficulty: 2 Status: Pipeline
2015-09-29_06.22.53

How we built a single GPU Linux workstation for deep learning.

We'll walk you through the process of building your own Linux computer optimized for Deep Learning. We recently got bit by the Deep Learning bug and after training a few Neural Nets on CPUs it became clear, we needed a GPU. One pivotal change in deep learning has been the advent of powerful GPU's capable of quickly performing the matrix computations necessary to train and test large neural nets. We base our workstation on Nvidia's GeForce GTX Titan X GPU. This GPU has 12GB of memory and a whopping 3072 cores. It comes highly recommended in Tim Dettmers excellent and exhaustive evaluation of different GPUs for use in deep learning. We use the newest Intel i7 processor, a the Hero Maximus motherboard from Asus, 16GB RAM and 1TB hard drive. There are a bunch of great sites for building a PC from scratch, we especially liked PC Part Picker and made use of their system build feature. Below we show a step by step guide to building the actual physical machine, which takes a few hours and is a lot of fun. We'll detail setting up the software environment for your Linux box in another tutorial to come.

Look at the GPU Glow

2015-09-29_06.56.08
The GeForce Titan X GPU glows green against the blue LEDs of the Sentey computer case.

Parts

Title Description # Cost Link Picture
NVIDIA GTX Titan X GPU (distributed by ASUS) ASUS GeForce GTX TITAN X Graphics Card, 12GB GDDR5 384-Bit, PCI Express 3.0 HDCP Ready SLI Support Video Card (GTXTITANX-12GD5) 1 $999.99 Link 51oipwolfrl._sx180_
Intel i7 CPU Intel Core BX80646I74790K i7-4790K Processor (8M Cache, up to 4.40 GHz) 1 $339.99 Link 51hdz_xpcil._sy180_
Motherboard ASUS Hero for Intel Asus MAXIMUS VII HERO LGA 1150 Intel Z97 HDMI SATA 6Gb/s USB 3.0 ATX Intel Motherboard 1 $209.99 Link 415m_kt3yul._sx180_
Power Supply Bronze EVGA EVGA SuperNOVA 750 B2 80+ BRONZE, 750W Semi Modular NVIDIA SLI and Crossfire Ready 5 Year Warranty Power Supply 110-B2-0750-VR Value: 750W 1 $85.99 Link 51wulazrmql._sy180_
RAM Crucial Ballistix Sport 16GB Kit (8GBx2) DDR3 1600 MT/s (PC3-12800) UDIMM Memory BLS2KIT8G3D1609DS1S00/ BLS2CP8G3D1609DS1S00 Value: 16GB 1 $66.99 Link 410h5n2tsyl._sx180_
Desktop Enclosure Standard ATX Sentey® Gaming Computer Case Gs-6011 Blade / Transparent Side Panel / 1 X USB 3.0 + 1 X USB 2.0 + Hd Audio/mic / Removable Bays 3.5 and 2.5 / 2 X 12cm 1 $54.99 Link 51gjiyrtxdl-1._sy180_
Magnetic Hard Drive WD Blue 1TB Desktop Hard Disk Drive - 7200 RPM SATA 6 Gb/s 64MB Cache 3.5 Inch - WD10EZEX Value: 1TB 1 $53.99 Link 51fblqlvjfl._sy180_
WiFi PCi Card TP-LINK TL-WDN4800 Dual Band Wireless N900 PCI Express Adapter,2.4GHz 450Mbps/5Ghz 450Mbps, Include Low-profile Bracket 1 $37.24 Link 41dj0fwrysl._sy180_
CPU Fan Cooler Master Hyper 212 EVO - CPU Cooler with 120mm PWM Fan (RR-212E-20PK-R2) 1 $31.24 Link 518bexkeiil._sy180_
Permalink: http://lucidtronix.com/tutorials/68
Quick tutorial on how we setup a Raspberry PI from scratch....
The MCP79410 breakout board gives easy access to this cute little timekeeper....
Wall-ter is a wall following robot....
Rainbow flower necklace using the TLC5940....
Play pong on a little LCD screen with two big joysticks....
Control a 120v outlet with a 5V arduino....
From simple rules emerge mysterious and intricate patterns....
Make an interactive light show using a proximity sensor, and a bunsh of shift registers and LEDs....