Over the years, many corporations – and home users – accumulate discarded computers, printers, and laptops. If nobody disposes of them immediately as electronic waste, they often stand around unused in storage rooms. A larger market for used equipment has established itself, and used systems often find buyers or lessors quickly. Because of tax regulations and leasing contract terms, the market is fed by company inventories with equipment only a few years old.

Prices drop extremely quickly, especially in the case of professional computer systems. Moreover, new owners might be able to write these devices off immediately for tax purposes as low-value assets for an instant benefit. The new owner also contributes to a positive ecological balance, because the production of new computer systems is resource intensive.

Unlike new versions of Windows, which might not provide drivers for older hardware, Linux distributions potentially support older hardware better than brand-new equipment. Some special Linux distributions with the right drivers concentrate specifically on such older hardware, but which used systems can admins best buy and run with Linux? Which desktops support production use with the lowest possible administration overhead?

Hardware Criteria

When you are choosing used hardware to run Linux, you need to consider several things up front to prevent imminent failure or costly upgrades of the systems after a short service life.

For example, it is a good idea to look primarily for older high-performance systems from known sources: Conventional desktop PCs for all-around use in offices quickly reach their performance limits or require more RAM or larger hard disks. Because most contain mid-range components, they are also not suitable for computationally intensive tasks you can find in CAD or multimedia environments.

Professional workstations or servers are far better suited candidates for repurposed systems, and they usually have well-equipped basic configurations. In the mobile segment, you should look for the business series in used laptops. In these cases, manufacturers usually offer replacement parts such as rechargeable batteries and optical drives for a considerably longer period of time. Additionally, business devices are generally built with far more valuable components than consumer goods, and because they are more solidly constructed, manufacturers often offer longer warranty periods.

Some pricy portable computers offering workstation performance are consistently tuned by manufacturers for performance. In addition to powerful processors and large memory capacities, they also come with high-performance graphics units, allowing the computers to cope easily with graphically intensive applications. IT buyers purchasing used mobile systems need to differentiate between business all-arounders and mobile workstations, depending on the application scenario.

Software Criteria

A suitable Linux distribution, in terms of available hardware capacity and expected performance potential, will ideally allow you to manage systems easily and give you access to as extensive a software repository as possible, which ensures that users have a sufficient number of programs from which to choose.

In the enterprise environment, the systems should also be easy to use. Because of the many working environments under Linux, you will have to make your selection carefully: On computers with relatively little RAM, resource-consuming desktop environments such as KDE Plasma or Gnome are of little use, and even Xfce can be out of place, because desktop environments can occupy more than 500MB of RAM from the outset.

If you run several large standard applications (e.g., LibreOffice, Gimp, Firefox) simultaneously, you will stress the swap partition. A slow hard disk by today's standards will make every task a test of patience.

Significantly fewer resources are consumed by lean desktops like Mate, LXDE, and LXQt. Although they lack detailed configuration options, they are generally based on operating concepts that have been around for decades and require little training. These desktops usually leave enough free memory for heavyweight applications, even on lower powered computer systems, so they rarely or never use the slow swap partition.

Test Scenario

For the practical tests, several older computer systems from different generations were used. In addition to a 2005 HP Compaq NC6220 laptop with a single-core Dothan series processor, I tested a more modern Fujitsu Siemens notebook with a Core 2 Duo processor from 2008 and a high-performance HP Z600 workstation with two quad-core Xeon processors from 2010.

The latest system to be tested was a 2013 HP Elitebook 2570p with a third-generation Core i5 CPU. With the exception of the two oldest notebooks, all devices were equipped with SATA SSD mass storage. In the older systems, I also upgraded main memory to a mid-range configuration; the oldest HP Compaq notebook had a minimal 2GB of RAM.

Out of competition, but for comparison, I also tested the third-generation Raspberry Pi microcomputer (built in 2016). Above all, I wanted the small-board computer (SBC) to give me a feeling for the hardware performance of the veteran computers and show whether it makes sense as an inexpensive desktop replacement for simple office workstations.

The lean test distribution was the Q4OS Debian derivative [1]. Because this operating system is available in both 32- and 64-bit versions, the test results are easy to compare. After a fresh installation, the test team measured the startup behavior on all computers using the appropriate system tools and then determined the memory requirements without any applications running. Under Q4OS, especially, the various desktop environments were tested because the Debian descendant can be installed and used with just a few mouse clicks.

Another test checked processor performance for demanding tasks (e.g., transcoding a DVD with HandBrake [2]), always launching the software with the same settings.

Q4OS was also installed on the Raspberry Pi so it could be compared directly with the full-grown systems. With the exception of video transcoding, it solved the same tasks as Intel-based systems with the same parameters, making the differences in computing power clearly visible and throwing light on the tasks for which the SBC is better suited in everyday office life and how close its performance comes to Intel-based systems.

The idea behind the test was not to collect synthetic standard benchmark results. Benchmarks usually only measure the performance of individual components. When it comes to the real application performance of a complete system, benchmarks have little informative value.