In this post we are going to learn about Venus, my deep learning computer, and how I built it. Along the way, I will explain at a high-level what each hardware component of a computer does and how I navigated the landscape of selecting parts for a functional build. I'll also describe how I installed relevant software for the machine and include some benchmarks showing the superior performance of a GPU system over a pure CPU system. WARNING: this is a pretty long post that functions as a complete tutorial for building a deep learning computer literally from scratch, no assumptions made. But…since it's long I highly encourage you to peruse and skip any sections depending on your interest. While there are numerous build descriptions out there showing how people constructed their own deep learning rigs, as I went about consulting some of them, I often felt there was some crucial component missing. As you start on your build journey, it's easy to get mired in the weeds of hardware terminology. Should I pick an M.2 SSD or will SATA suffice? Can I get away with HDD? How many PCIe x16 slots do I need? Should I pick DDR4-3000 or DDR4-2400 memory? All this lingo can be very overwhelming especially for newcomers to hardware. But before we start shamelessly name-dropping so that we sound smart, let's go back to the fundamentals.

The machine I built costs $3k and has the parts shown below. There's one 1080 Ti GPU to start (you can just as easily use the new 2080 Ti for Machine Learning at $500 more -- just be careful to get one with a blower fan design), a 12 Core CPU, 64GB RAM, and 1TB M.2 SSD. You can add three more GPUs easily for a total of four. Assuming your 1 GPU machine depreciates to $0 in 3 years (very conservative), the chart below shows that if you use it for up to 1 year, it'll be 10x cheaper, including costs for electricity. Amazon discounts pricing if you have a multi-year contract, so the advantage is 4–6x for multi-year contracts.