I finally decided to invest in a pc that could handle a few deep learning projects. I know I could always use AWS, but their is something to local processing that simplifies things. And also, I am totally dual-purposing and mining zclassic and etheruem on the side.
Building your own computer may sound intimidating, but nowadays with websites like pcpartpicker.com, you can quickly learn from other builds, identify parts that fit your needs, and save time on shopping with quick comparisons on lowest prices. Compliment this with a quick youtube search on “live pc builds” and the magic of amazon prime shipping…you can be starting your pc project by the end of the week. Here we go…
With Deep Learning and Mining in mind, I needed to invest in a few good GPUs, luckily I did this around cyber monday and saved a few hundred bucks before the Nvidia price markups. I spent $1187 in total on this pc, which includes my initial GeForce GTX 1060 6GB GPU card I picked up for $287 from newegg. Here is the final part list here.
After understanding my needs, here was my process of part selection…
- 1 ) Pick Your Motherboard. I choose a board with plenty of room to upgrade RAM, have multiple GPUs (up to 4), Wifi enabled, easy to use BIOS, and good price.
- 2) CPU vs. GPU. I once read that CPU is like a sportcar and GPU is like a dump truck. This analogy has stuck with me when explaining the purpose of both. CPU can perform tasks very quickly and GPUs can process heavy loads (basically matrix multiplication which supports crypto-mining and Deep learning algorithms). So if you are building a pure crypto miner, having a cheap CPU is just fine because that rig will just run 24/7 on same miner software. If you plan on using the pc for web-browsing, and general use with multiple applications…having a good CPU is critical. I went with above average CPU and very respectable GPU.
- 3) Memory – I decided to go with 16 GB, which is enough for my purposes, and have room to upgrade if needed.
- 4) Power Supply – This PS may seem like an overkill with 850 Watts, but I didn’t want there to be an issues if I actually load up with 4 GPUs eventually.
- 5) Case – I wasn’t concerned about space, and wanted room to expand for years to come. This case looks clean and is a joy to work in. It also has 2 built-in fans, and nice plexi-glass window.
- 6) Cooling – Standard CPU cooling setup for tower builds.
- 7) Operating System – I decided to go open source and am running on Ubuntu 16.04.
This was an absolutely satisfying project and probably benefited more from the “process” of building than the end product. Which if you think about it…is a great way to live. Enjoying the journey once again.