How We Test Laptops (2026 Update)

It takes us three days to test a modern laptop. Our testing methodology is exhaustive and has been honed and tweaked over the past three years so buyers can easily see if a potential purchase will meet their needs. It’s also led to a scoring system that can accurately compare lightweight ultraportables with big, chunky gaming laptops to produce an authoritative, overall Best Laptop List. The latter gets updated every month, along with group tests for portable, business and gaming notebooks. What follows are details of what the scores are and how we reach them.

We also explain the pre-test preparation each model goes through which buyers of new laptops may find useful. The following has all been updated for 2026.

The Short Version

Here’s the tl;dr for how we assign points. Each section is then mathematically converted into a 5-star rating where 1 is the weakest score we’ve seen in the current (and previous) calendar year and 5 is the best:

Features, Ergonomics and Design – Focuses on what the laptop can do and how it handles.

General Computing & CPU Performance – We run benchmarks and tests that focus on the main processor (CPU) and the system as a whole.

Graphics and Gaming Performance – We run tests and gaming benchmarks that focus on a laptop’s GPU performance.

Portability and Battery Life – We assign scores for Battery Life, Weight (including cables and power brick) and Build Quality.

Price and Value – We assign scores depending on how cheap or expensive a laptop is. The Value score says whether it’s worth it or not. These scores (in reviews) are initially based upon the launch RRP but change every time we do a group test according to price drops and rises.

Overall Score – The final score reflects the laptop’s overall rank in the market.

The Long Version

How to set-up a brand new laptop for the first time

The following will be useful to all the people who go on social media and say things like, ‘I just bought a new laptop, now what?’ It will also help people who fire-up their brand new laptop only to discover that it’s loud, hot and very slow. Those people need to understand that it’s updating in the background. This will happen automatically and will take more-than three hours over a very fast internet connection. It will include numerous updates to Windows, drivers, vendor applications and often firmware. When the latter is updated it may very well kick-off another full cycle of updates. We force this to occur manually until we’re satisfied that all updates have been completed. Even then, you can expect something else to update in the background.

Updating, Disabling and Removing Software on a new laptop

The first thing we do is ensure that everything is up to date and that any software that could interfere with testing is uninstalled or disabled. This means running Windows Update until every new update, including Optional Updates, is installed. We update any vendor apps that come with the laptop. We then run the vendor apps to see if they pick up updates that Windows didn’t (this is getting increasingly rare, but still often affects things like drivers for screens, audio and keyboards). We use Nvidia and AMD software for updating the graphics drivers (downloading graphics drivers independently can cause major issues).

To stop tests being impeded while running, we uninstall any third-party antivirus software (usually McAfee). We don’t recommend that you do that, though – make an informed choice, yourself. We make sure tools like OneDrive have been signed-out and that any start-up apps, that don’t need to be preloaded, are disabled in Windows’ Start-up options – this usually means OneDrive, Adobe Creative Cloud and Steam among others.

Once we have a relatively clean Windows Desktop that’s not constantly trying to update itself in the background, we install and run our benchmarks. Note, even when you think a laptop is updated, it often finds a way of updating something else. A few reboots usually gets this out of the way. It mainly happens with vendor apps and after Firmware updates which can trigger a whole slew of additional Windows and Driver updates.

It’s worth dispelling a myth for anyone with a new laptop reading this – we absolutely DO NOT RECOMMEND reinstalling a ‘clean’ version of Windows to remove bloatware. Those days are long-gone. You will very likely break your laptop. Those apps and drivers that vendors supply often contain essential software for your laptop to function.

Features, Ergonomics & Design Scores

Below is a list of criteria and the maximum number of points they can score for each:

Performance Testing

We used to separate testing into 2D and 3D components but, in reality, multiple components of a system are used in many benchmarks. The appearance of Neural Processing Units (NPUs), for AI, further muddied the waters. We’re formally adding new benchmarks for 2026 but, in reality, we’ve been testing models using them for many months already.

General Computing & CPU Performance Scores

PCMark 10 – Removed for 2026

This stalwart just goes wrong way too often nowadays. It also doesn’t (fully) work on Snapdragon or Apple laptops whereas Geekbench does.

Geekbench 6 CPU – Added for 2026

2026 Update. Regular readers will know that we’ve been citing Geekbench scores for many months already. Now that we’re regularly reviewing SnapDragon and Apple laptops, it makes sense to add the Geekbench 6 cross-platform benchmark to the mix. We use the Multi-Core score for our overall results but list the Single-Core score (which is more for technical interest than real-world impact) for reference, too.

Cinebench Rendering Tests – Focusing on Cinebench R24 in 2026.

Cinebench R24 is replacing Cinebench R23 and R15 moving forward. There’s little benefit from running all three and R15 doesn’t work on across all platforms anyway.

UL Procyon Office Productivity – Removed for 2026

Another buggy mess which kept breaking due to driver and application updates. Microsoft’s regular deprecation of features meant scores were becoming less and less comparable every month. We also haven’t seen a laptop that can’t capably run office tasks for many years now.

UL Procyon Photo Editing – Removed for 2026

We wanted to keep this as it utilises the popular Adobe Photoshop. However, it too has become a broken mess like its Office-based siblings and it’s become much more hassle than it’s worth. It also relies upon Adobe’s LightRoom Classic which isn’t popular and breaks the benchmark regularly. It may return in the future but, in reality, all modern laptops can do general Photoshop tasks and only extreme power users might have raised an eyebrow at a high performance laptop taking 15 minutes to perform a demanding,, automated workload while a budget model (that they’d not likely buy) would take 30 minutes. Geekbench caters for photo editing anyway.

UL Procyon Ai Computer Vision Benchmark (CPU) – Removed for 2026

Laptop processor manufacturers love to talk about this benchmark but, despite having listed to every lecture on the subject, we still couldn’t make a case for buying a laptop over another because of the score churned out by this test. It’s more of a feature of CPUs than something like a dedicated GPU.

Graphics & Gaming Performance Scores

The following tests all rely heavily on a laptop’s GPU which is either embedded within the CPU or discreet – meaning it’s a separate chip (think, Nvidia). We always favour using the latter for testing, if present.

GPU-based workloads

UL Procyon Ai Computer Vision Benchmark (GPU) – Removed for 2026

Same reason for the CPU version.

UL Procyon Video Editing Benchmark

We always favour the GPU version of this test because performance is much higher than with a CPU. It’s one of the biggest ways to differentiate laptop power, in the market, and better performance will save video editors a great deal of time. You can find more details about what’s involved, here.

To help put these scores into perspective, the Nvidia RTX-4070-wielding Asus ProArt PX13 scored 27,673 (score: 3.2 / 5). Conversely, an Acer TravelMate Spin business laptop, with a modest, Intel Core Ultra 7 155U, 12-core Hyperthreaded processor and integrated Intel Xe graphics, scored 6,905 (1.4 / 5) and completed the H.265, 4K video edit-and-export benchmark in 306 seconds. The PX13 did it in 90 seconds.

Geekbench 6 GPU

2026 Update. The Geekbench 6 cross-platform benchmark is now added.

Gaming tests

For consistency, we focus on UL’s 3DMark gaming tests to mathematically derive scores but we occasionally post full-game benchmarks for reference.

3DMark

With 3DMark, we start with the hardest tests which only work on x86 PCs – Speed Way and Port Royal. These represent the latest-and-greatest games with the most realistic, eye-candy settings. They run at a 1,440p resolution with difficult ray-tracing settings turned on. They require the most powerful gaming hardware to run at smooth speeds.

In 2026 we’re also now focusing on UL’s most-modern 3DMark tests as their scores are more-consistent on modern laptops across multiple platforms.

The Steel Nomad, multiplatform gaming tests, ape modern, 4K-resolution games without using ray-tracing. For reference, the Asus ProArt PX13 scored 2,320 (average 23.2fps) in the primary benchmark (score: 3.1 / 5). In the lesser, 1,440p, “Light” version of the test, it scored 10,873 (average 80.5fps) which is a score of 3.7 / 5.

The easier, 1,440p, multi-platform, ray-tracing Solar Bay benchmark, is targeted at ‘mobile’ laptops. As an example, the PX13 scored 44,401 which is a healthy average of 169fps and represents a score of 2.8 / 5. For comparison, the long-battery-life, ultraportable Lunar Lake-based Asus Zenbook S 14 scored 15,228 (score: 1.4 / 5) which is an average of 58fps while the hefty, gaming laptop, the MSI Vector 16 HX, scored 92,286 (score: 5 / 5) which is an average of 350fps.

We also run some of 3DMark’s legacy tests to compare with older models and with cross-platform laptops that can run some of these but not others. None of them use ray-tracing.

Time Spy and Fire Strike Extreme are removed for 2026 as UL has already replaced them with the above tests which scale much more predictably and reliably.

We’re keeping Wild Life Extreme because it works across all platforms. Night Raid will be mentioned in some reviews just to illustrate whether a low-powered laptop can run basic, Full HD games. It scales horribly, though, and won’t factor in overall results any more.

Portability & Battery Life

Weight

Nobody likes to lump a huge, desktop replacement gaming laptop around, but that’s a compromise when compared to less-powerful, ultraportable laptops. So, while the latter often score poorly for gaming, here’s where they make a comeback.

We physically weigh the laptop, as manufacturer-stated weights are often wrong (or refer to a different SKU to our review unit). We also weigh the charger and cables as these can vary in size dramatically – some have phone-style chargers while others have enormous power bricks that weigh over a kilo.

Battery Life

For battery tests, we turn the screen down to the laptop’s lowest viewable setting in a dark room (or night-flight airline cabin). We don’t worry about calibrating a screen’s brightness for consistency because many laptops’ screens simply don’t get very dim – and that’s a differentiating feature.

Where possible, we run PCMark 10’s Modern Office Battery Life benchmark which runs the office features of PCMark 10 until the laptop dies. It only works on x86 laptops, though.

We also run the Battery Life version of UL’s Procyon Microsoft Office benchmark.

For laptops that don’t run either of these tests, we run lesser tests like UL’s Video Playback benchmark to give us some idea of battery life, but it’s not directly comparable.

It’s worth stressing that your mileage will vary depending on how you use your laptop and what you use it for.

Build Quality

We also also look at how sturdy a laptop is and give it a score out of 5. This primarily means checking to see whether it’s robust enough to survive life on the road. We also factor-in the strength of the hinge, which is a common failure point for laptops. Because of the non-scientific measurement here, it’s not weighted heavily.

Price & Value For Money – 2026 Update

Price is one of the most important factors when buying a laptop and it can vary massively. That’s why it’s got it’s own score and why a cheap, well-featured laptop can compete with an expensive, all-singing-all-dancing rival.

Our Value For Money score is derived by adding up 25% of each of the above category scores and multiplying it by an indexed price figure from across the market.

In 2026, we’ll be adding prices from other countries because overseas readers can’t easily visualise what our native Aussie dollar represents.

Overall Score

Our overall score is not an average of the other scores: we tot-up all of the other scores and the lowest total scores 1 / 5 while the highest total scores 5 / 5. We do this so that the Best Laptop Award Winner scores 5 / 5 as it looks strange if the top dog averages something like 3.3 / 5.

We also have awards for:

Best Gaming Laptop – the best model (using the above criteria) that also has a fast screen and discreet GPU.

Best Budget Gaming Laptop – As above but with a ceiling price of $2,000. That threshold rose considerably during COVID and, unfortunately, hasn’t fully come down yet.

Best Portable Laptop – the laptop with the highest Portability score, irrespective of price.

Best Business Laptop – this has some unique criteria based upon business requirements which are detailed in the group test.

Best Creator Laptop – These laptops have discreet graphics for media professionals and high-quality screens with design-oriented certifications.

We also give out Recommended Awards (if a laptop scores 4 or over in a pertinent area) and Highly Recommended Awards (if it scores 4.7 or over) in each category.

All of the above have been tweaked, tuned and iterated upon over the past three+ years. It’s an ongoing process and we expect them to be tweaked further over time.

Last Updated on March 1, 2026 by Nick Ross