Hardware

We want our hardware to keep doing good work after it’s cycled out of active use in our fleet. So we distribute rebuilt machines and/or components to people who can give them a good home.

This is the development diary for Greenhouse 2, the second iteration of my homebrew micro-rack.

2020-05-01 – Introduction

The original Greenhouse was a success, but several lessons were learned from it.

• Use of mITX boards minimized footprint, but drove up cost and reduced choice
• Over/under PSU design minimized horizontal footprint, but led to increased vertical spacing and more complex mounting
• Minimizing horizontal footprint wasn’t worth it
• Fully open design probably didn’t cool as effectively; pushing air where you want it isn’t as effective as forcing it to be pulled through where you need it
• Even moderately complex cutting and shaping of metal is a nightmare without proper tooling and space

And so the next iteration will be changed in several fundamental ways.

I’ve been working on adding support for the ARM architecture to Homefarm for a bit now. This week I decided to try to push through, and get that work done and tested.

Obviously you can’t test software designed to manage a farm of computers with just one machine, so I had to build myself a tiny ARM farm.

Building the cube

Here’s most of the raw materials:

The Firepear computing stack is currently all Ryzen. It’s simple cost-benefit analysis: I’m interested in crunching as much data as possible per unit time, at a reasonable cost (in both money and electricity). Right now, that’s Ryzen.

I’m upgrading all my machines to the new R9 3900X CPU. This provides an opportunity to compare all three generations of the Ryzen family.

3900X vs 3950X

I want to throw as many cores as I can at the problems that I volunteer compute time for, but I also have a finite budget. The 3950X will be beyond my reach at launch, so my strategy is to upgrade to 3900X starting in early July.

After months of pondering what GPU to use to get my compute farm back into crunching GPGPU workunits, I was persuaded by a recent review to order a GTX 1650 and give it a go.

Background and lead-up

Years ago, I was reading stories about how Linux support on Steam had really improved. I decided to build a Linux box capable of running modern-ish games. This was in 2015, a time when AMD GPU support on Linux was abominable, so my choice for a video card boiled down to “which Nvidia card?” My answer was the GTX 750 Ti.

This is the development diary for Greenhouse, my attempt to build a DIY compute farm in a box. In other words, a multi-machine chassis.

2018-11-01 — Introduction and Plan

The core idea is to assemble individual compute nodes on trays, in the most compact way feasible, and then to stack nodes atop each other in an enclosure.

The components of each node will be:

• A tray made from some thin, rigid, wood-based board.
• Nylon washers/spacers and steel bolts, for standoffs and attachment of mobo to tray
• A Mini-ITX motherboard with wifi and an M.2 slot
• A CPU
• 2x4G RAM
• An M.2 SSD (cheap now, and eliminates 2 cables)
• An SFX PSU
• 2x 120mm fans in a →[fan→mobo→fan]→ configuration

This will be prototyped out using cardboard in place of a hard tray, to determine optimal placement and connection of components. Already established is that: