Linux on an old laptop

A client brought me a laptop last spring with the words: "the computer shop said it's finished, it won't take the new Windows." It was a ThinkPad from 2012. Core i5, 8 GB of RAM, a mechanical drive that sounded like it had opinions. The shop was right about one thing and wrong about everything else.

Right: that machine will not run Windows 11. Wrong: that this makes it finished. I hear the sentence "it can't be done" often enough that it has become a signal for me rather than a conclusion. Usually it means one of two things — the person saying it does not know how, or they do not feel like it. Both are fixable. The hardware rarely is the problem.

So I did what I always do instead of arguing: I measured. Stopwatch, the same laptop, three states — before, after the parts, after the system. Below are the numbers and, more usefully, which of those changes actually did the work.

Why an old laptop feels old

Here is the thing almost nobody tells you at the counter: a computer that has become unbearably slow is usually not slow because of its processor. A Core i5 from last year is genuinely faster than one from 2012 — but perhaps two or three times on a single task, not the twenty times the difference in experience suggests. For opening a spreadsheet, the drive decides more than either processor does. Office work stopped being processor-bound somewhere around 2010. What changed since then is everything around the processor.

Three things make an old machine crawl, in this order of impact:

• A mechanical hard drive. This is the one. A spinning platter needs several milliseconds to reach the data; an SSD needs a fraction of one. Everything a computer does — booting, opening a program, saving a file — is thousands of those trips.
• Not enough memory. When RAM runs out, the system starts shuffling data onto the drive. On a mechanical drive that turns slow into unusable.
• Heat. Dried thermal paste and a fan packed with dust mean the processor overheats and deliberately slows itself down to survive. It is protecting itself, and you experience it as the machine getting old.

Notice that none of the three is the operating system. This matters, because it sets a realistic expectation: swapping Windows for Linux on a machine with a dying mechanical drive will give you a slightly faster disappointment. The system is the last step, not the first.

What I actually did with that ThinkPad

In order, and with the stopwatch running at each stage.

First, diagnosis rather than assumption. The drive's SMART data — the health report every drive keeps about itself — had reallocated sectors, which is a polite way of saying parts of the disk had already failed and been quietly worked around. That single reading decided the whole job. There was no point tuning anything on a drive that was in the process of dying.

Then the physical pass, which takes about forty minutes and which almost nobody does: open the case, remove the dust from the cooler, replace the thermal paste, reseat the memory. Boring, cheap, and it recovers more speed than most software tricks.

Then the parts: an SSD in place of the mechanical drive, and memory topped up. On a machine of that age both are inexpensive, and both are things the original manufacturer intended you to be able to change — which is itself a good argument for buying business-class hardware second hand.

Only at the very end, Linux Mint. And here is the honest accounting of what each step contributed: the SSD did most of the work, the thermal paste and memory did a good part of the rest, and the system did the remainder — mostly by not consuming a couple of gigabytes of memory before you have opened anything.

Why Mint specifically, and not one of the dozens of other options? Because for someone coming from Windows the desktop is familiar enough that they do not have to relearn anything: a start menu bottom left, a taskbar, windows that behave like windows. It is built on Ubuntu LTS, which means long support and a huge amount of software packaged and ready. For a machine going to a person who wants to work rather than tinker, that is the correct trade-off.

Where this genuinely does not work

I would be doing exactly what the computer shop did if I told you this always works. It does not, and the failure cases are predictable enough to check in ten minutes.

Software with no Linux version and no working equivalent is the main one. Polish accounting packages, some industry-specific tools, CAD with a hardware licence key, certain e-signature and banking tokens. Games with kernel-level anti-cheat will not run and the publishers are explicit about it. Very old peripherals with Windows-only drivers are a coin toss — a twelve-year-old laser printer usually works fine, a specialist label printer often does not.

There is also a hardware floor below which none of this helps. Two gigabytes of memory, a processor from before roughly 2010, a laptop whose battery and hinges are both gone — at that point you are restoring a machine out of sentiment, not economics, and I will tell a client that rather than take their money.

My working rule: list the programs the person genuinely cannot work without, then decide. Never the other way round. It costs a quarter of an hour and it prevents the one outcome nobody wants — a machine that boots beautifully and cannot do the job it was bought for.

What such a revival actually costs

People ask me for a price before they ask me for a diagnosis, which is backwards but understandable. So here is the honest shape of it, without quoting figures that would be out of date by the time you read this.

Two components dominate: an SSD and memory. Both are commodity parts, both are cheap for machines of this generation, and both are the reason the whole exercise makes financial sense. Everything else — thermal paste, a can of compressed air, a couple of hours — costs almost nothing in materials and quite a lot in attention, which is the part people underestimate.

Where the real time goes is not the screwdriver work. It is the data. Getting somebody's documents, photographs, mail archive and browser passwords off an old system and onto the new one without losing anything is the part that takes care, and it is the part where a rushed job shows. I do a full copy before touching anything, keep it until the client confirms everything is where it should be, and only then release the space. That habit has saved me twice. Both times were my own fault.

Compare that against the alternative and the decision usually makes itself. A new machine is not only its price — it is the migration anyway, the licences again, the peripherals that may or may not fit, and a week of somebody being slightly lost. The revived machine skips most of that, because to the person using it, it is still their computer.

What it is worth, in money and in landfill

The laptop went back to its owner for the price of an SSD, some memory, thermal paste and an afternoon. Against a new machine, the arithmetic is not close. And the replacement would not have been faster at the things that person actually does — writing, email, a browser, occasional video calls.

There is a second column in that spreadsheet that rarely gets filled in. Producing a new laptop costs a great deal of energy and materials before it is ever switched on — the manufacturing footprint dominates the lifetime one for a device used in an office. Keeping a working machine in service for another three or four years is, by a wide margin, the most effective environmental decision available to a small company. It also happens to be the cheapest, which is a rare and pleasant combination.

This is the part of my work I like most, and I say that fully aware that I earn money from it. A machine described as finished goes back to a desk and does another few years of honest work. Nothing about that requires belief in anything. It requires a screwdriver, a stopwatch and a refusal to accept "it can't be done" from someone who has not tried.