Building a Deep Learning GPU Server

Amazon Web Services (AWS) and Google Cloud Platform (GCP) can be phenomenal for processing deep learning models in the cloud, but there’s a catch - especially if you’re a student or enthusiast with no corporate account access - These cloud tools get expensive! For $1-3 per hour, you can cloud compute using AWS or GCP… Not bad! But if you want extended access to your cloud instances, or 247 access as I’d like to have, cloud solutions become cost prohibitive.

Alas, my recent obsession for a fun home project… Building my own deep learning GPU server that I’d keep at home and access remotely, as needed. I’ve never built a computer so this project has me excited, a little scared, and eager to run my personal models faster… This would help so much in the data science process of evaluating models and iterating / improving more quickly.

If you have words of caution or advice, or maybe something major that I’m missing, please shoot me an email or contact me via LinkedIn or GitHub!

Source:
Why building your own Deep Learning Computer is 10x cheaper than AWS by Jeff Chen


blog · scatter podcast · resources · contact · who is javier? · main