Methodology
How we rate, and what the labels mean
Read this before you trust a number on this site. Every score points at evidence you can check, the byline tells you whether a human ran the laptop, and where we don't know something we say so.
The seven scores
Seven categories, each scored 0 to 10, each pointing at something you can check. Linux readiness is the only one computed by formula; the other six are human judgements with cited sources.
| Score | What it measures | What defends the number |
|---|---|---|
| Linux readiness | Out-of-box behaviour and per-component support on the best documented distro. | Computed from the per-distro hardware table, not hand-set. Formula below. |
| Battery | Real-world runtime with WiFi on, not the manufacturer's lab figure. | Logged test runs or community reports, with the test conditions stated. |
| Display | Panel type, measured or reported nits, refresh rate, PWM behaviour. | A cited measurement, or a clearly labelled manufacturer claim. |
| Build | Chassis, hinge, flex, port selection, serviceability. | iFixit teardown scores, ports we counted, soldered components called out. |
| Keyboard and trackpad | Travel, layout quirks, trackpad size and driver behaviour. | The things you only notice after a week. First-hand on the Reviewed tier. |
| Performance | CPU and GPU class against what the machine is for. | Sustained-load benchmarks on battery, not just peak plugged-in scores. |
| Value | The score against the real price in your region. | Recomputed when the price moves. A great laptop at a bad price loses here. |
If a number has no defence, it does not ship. Sources are vendor specifications (labelled as claims, overridden by independent measurements where we have them); community data (ArchWiki, linux-hardware.org, distro forums, Phoronix); and the laptops I own and run, the only path to a Reviewed byline.
Reviewed versus Compiled
A Reviewed by Dennis page may make first-person claims (week-of-use suspend, chassis thermals, fan noise, screen tearing) and must carry a test date plus the OS, kernel, and distro used. The Reviewed scope is small and stated: the Razer Blade 16 (2024), the Redmibooks, and a few Chinese laptops.
A Compiled and edited by Dennis page covers everything else and most of the site. It may claim spec accuracy and summarise community reports, but must cite a source for every claim, carry a "last verified" date, and say "unknown" where the evidence is thin rather than guess. Pages are re-checked on a rolling quarterly cadence; a score is only as current as the date next to it.
The Linux Readiness Score, in full
Every laptop gets a Linux Readiness Score from 0 to 10, one decimal, computed the same way every time from the per-distro table on each model page. Same data in, same number out, so you can check the arithmetic. Two things go in.
First, the clean-install grade per distro: works (boots clean, all six boxes green or amber, no terminal needed); minor tweaks (one or two documented fixes that stick across updates); problematic (a core component broken or fragile, usable only with effort); or unknown (no reliable evidence, dropped from the math, neither pass nor fail). Second, six pieces of hardware that decide day-to-day happiness.
| Component | Weight | Why that weight |
|---|---|---|
| suspend | 0.28 | The biggest real-world Linux failure. Broken s2idle kills the battery in a bag. Hardest fault for a user to work around. |
| wifi | 0.26 | No wifi is close to no laptop, and you cannot even download the fix. Edged out by suspend only because dongles and tethering exist. |
| audio | 0.16 | Speakers and mic. Blocks calls, but degrades to USB or Bluetooth audio. Serious, not fatal. |
| webcam | 0.14 | Blocking for many, but the most substitutable: a 20 euro USB cam fixes it. |
| bluetooth | 0.10 | Mouse, headphones, files. Annoying when dead, almost always a wired path. |
| fingerprint | 0.06 | A login convenience. Often absent by design on Linux-first machines, and password login always works. Lightest on purpose. |
The clean-install grade and the hardware split 45 / 55. Hardware gets the larger share because the six boxes are falsifiable; the grade still keeps a heavy 45 percent because it catches what the boxes miss (GPU driver pain, thermal throttling, an unusable trackpad), and neither term may fully override the other. That anchor is why a laptop with five of six boxes green but a problematic grade, the Dell XPS 13 with a dead webcam, does not float up to a 9.
We score the machine at its best documented distro and name it; averaging in a distro nobody tested would only punish it for the size of our test bench. Unknown never helps the score. If we know too little to score fairly (the bar is a known clean-install grade plus at least three of six known components) we print "Not rated". You will see that on the MacBook Air M4: Linux there means Asahi, and Asahi barely boots on M4.
Next to every score is a confidence level (high, medium, or low), a separate number telling you how much we know, not how good the machine is. NVIDIA laptops are capped at medium even with complete data, because the proprietary driver stack changes under you and a three-month-old score may not survive the next kernel. Where an upstream tracker is the live truth, such as Asahi, we link it and you should trust that over us. If you only need wifi and suspend, read the per-distro table, not the headline.
Staging, and disagreeing with a number
New long-tail pages can start un-indexed until they have enough depth, then flip to indexable. This is staging, not cloaking: the same content is shown to every visitor and crawler; only whether we ask search engines to list a thin page changes.
Disagree with a number? Good. Every score points at a source and the Linux formula is public, so it is an argument about evidence, not taste. Tell us where the arithmetic or source is wrong via the contact page and we will fix it or show our working.
About and impressum Review program How we intend to make money Privacy