We test so you don't waste money — here's how we review laptops, tablets and PCs

Tom's Guide

3 days ago

Buying a computer is a big decision, and it can be a difficult one. There's often no easy way to know from an ad or a manufacturer's website what makes one PC superior to another. To make matters more confusing, two PCs could have nearly identical specs, but one could still be a better buy for reasons you've never even considered.
That's a big part of why we do what we do here at Tom's Guide. We've been there, and we want to save you from the headaches we've experienced when buying computers for ourselves. And with our knowledge and expertise, we can confidently point you towards the best laptops, gaming laptops, computers, gaming PCs, MacBooks, Chromebooks and tablets you can buy right now.
The process all starts with lab testing — not merely in terms of how quickly you might edit photos or last while on battery, but also in putting them through the same paces you would.
We dig deep to get the numbers that tell the performance story, and then we dig in, spending real time with them browsing the Web, sending e-mails, playing games, and so on until we have a thorough picture of what it means to live with the computer. Whether you're paying $500 or $5,000, that's what matters most.
Here's every single quantitative and qualitative test that we subject every device to — so you'll know exactly what you're getting when you buy one and hold it in yours.
The vast majority of systems we test run Windows, which makes sense as Windows PCs account for about 75% of all computers used. So many (but not all) of our tests are Windows-centric and require some tweaks to keep the playing field as level as possible. But this helps ensure we give you the most — and best — information we can about every PC we touch.
We begin our testing process by turning off all of Windows' screen and sleep settings; some of our tests run for a long time and we don't want them interrupted, and in our experience testing is better performed 'warm' anyway.
Get instant access to breaking news, the hottest reviews, great deals and helpful tips.
We then set the power mode to 'Best Performance' so we're seeing the most of what the system, as configured, can do; we update all the system's drivers to the latest versions for the same reason. We always run all tests at the PC's default resolution while no other programs are open.
One important thing we don't do: fool around with any of the manufacturer's default settings. You deserve to know exactly what you're going to get — and how well it will run out of the box, so we don't try to boost performance by using any software or settings that aren't already in place when the PC boots up for the first time.
And if we discover that the manufacturer has changed any settings from what you'll get if you buy the system, we change them back. Your experience is the most important thing to us.
As exciting as we wish our lives were, and as much money as some of us (ahem) might spend on top-end hardware to max out performance in every metric, the fact remains: Most of what most of us do on our computers every day is fairly mundane.
Important? Yes, but far from earth shattering. Our benchmark tests reflect this reality, focusing on the tasks and applications that consume the greatest amounts of our time.
Copying files might be the second most basic organizational task you can perform on your PC. (The first, of course, is rearranging icons on your desktop.)
To see how quickly our test systems can do this, we run a script that copies a 25GB folder containing a variety of file types (including Microsoft Word documents; folders stuffed full of photo, audio, and video files; and a selection of application installers) to another location on the system drive.
The script calculates how long this takes, and we can also use that time to determine the speed (in MBps) of the transfer.
Primate Labs is renowned in the industry for its wide-ranging tests that assess key real-world performance metrics.
The Geekbench CPU test processes a lengthy series of cryptography, compression, rendering, image editing, speech recognition, machine learning workloads and much more — returning a simple, clear result that allows for comparisons with other devices running not just Windows but also macOS, Linux, Chrome OS, Android, and iOS.
The current version of Geekbench (6.4, as of this writing) calculates its scores using the performance of a Dell Precision 3460 desktop with an Intel Core i7-12700 processor as a baseline. Higher scores are better, and double the score means double the performance.
With AI now such a critical component of both hardware and software, it's no surprise that Primate Labs developed a test specifically to measure it.
Geekbench AI is similar to its namesake in scope and style, but targets nearly a dozen computer vision and natural language processing workloads and scores according to Single Precision (FP32), Half Precision (FP16), and Quantized (INT8) methods. The hardware baseline here is a Lenovo ThinkStation P340 with a Core i7-10700 processor.
Gone are the days when you'd need a fancy (and potentially expensive) program to convert video between formats — the free HandBrake does just that and a whole lot more.
We use HandBrake to convert the 12-minute 4K Blender film Tears of Steel from a 6GB MOV file to a 1080p30 MP4, determining the amount of time this takes with the built-in activity log. We report the result in minutes and seconds.
The Business Applications Performance Corporation, lovingly known in the biz as BAPCo, has been a driving force in computing for more than three decades — uniting many disparate companies (including, at present, Acer, Dell, HP, Intel, Lenovo, Samsung, and Western Digital) to develop industry-standard benchmark tests.
BAPCo's latest, CrossMark, currently runs on everything except Chrome OS, and measures performance in real-world productivity (document and spreadsheet editing, Web browsing), creativity (photo editing and organization, video editing), and responsiveness (application launching and file opening) scenarios. The test results in scores for each of these categories, plus an aggregate (which weighs the productivity and creativity tests higher than responsiveness).
Futuremark was a major PC benchmarking force before UL bought it in 2014, and things have only improved since then. One of the company's longest-lasting tests is 3DMark, a sweeping collection of tests for measuring all aspects of your computer's graphics performance.
For non-gaming computers, we keep it light, running The older Fire Strike (tuned to DirectX 11, or DX11, gaming) and Time Spy (DX12) on everything, as well as the newest all-purpose test, Steel Nomad. Systems without discrete video cards also get a round of Night Raid, which is optimized for testing integrated graphics.
Gaming has been a key computer activity since the industry's earliest days. But, for better or worse, most of us have progressed past the simple-but-addictive likes of Microsoft Minesweeper and Solitaire. We use a popular, but not-too-demanding game to see how well non-gaming systems can hack digital entertainment.
Our current title of choice is the habit-forming Sid Meier's Civilization VI, with the Gathering Storm update applied. With the graphics settings adjusted to medium values, to challenge less-powerful systems while not taxing higher-end machines too heavily, we run the built-in Gathering Storm Graphics benchmark and report the frame rate the game determines.
Though it's less of a problem today than it used to be, laptops can still get warm (and thus uncomfortable to touch or put on your lap) if they're used too heavily for too long.
After turning off the computer and letting it cool completely, and then restarting it and allowing it to idle for a few minutes to complete its boot processes, we play a 4K video, full screen, for 15 minutes.
After that time, we use an infrared thermometer to measure the chassis's temperature at the center of the touchpad, between the G and H keys (roughly at the center of the keyboard), and the center of the underside, and then 'sweep' the computer to find what part of it is hottest. Because games can be even more demanding, we have a separate test that measures the heat impact of those (see below).
No matter what kind of computer you have, you're using its display more than any other part (with the possible exception of the keyboard). So if you're shopping for a laptop or an all-in-one, where the screen is a key part of the experience, you deserve to know how good it is.
Color: We use a high-end Klein K10-A colorimeter and DisplayCal software to measure two key display traits: Delta-E (color accuracy — lower is better) and color gamut volume (how much of a color space the screen can show). We report results for both sRGB and DCI-P3 color spaces.
Brightness: Using the same K10-A tool and Klein's ChromaSurf software, we test display brightness at five points on the screen and average the results (in nits). For OLEDs, we adapt the test with a black background and a moving white window to account for brightness variation. If the device supports HDR, we also test brightness using 10%, 40%, and 100% screen patches.
Battery life is one of the most important features in any laptop, so we built a custom test to measure it consistently. Using Microsoft Edge on a closed Wi-Fi network, the system cycles through a set of web pages until the battery dies — reporting its status at regular intervals. Once recharged, we record the total runtime and run the test at least twice to confirm accuracy.
To keep results consistent across devices, we normalize each system: screen brightness is set to 150 nits, background apps (like antivirus) are disabled, and a custom power plan is applied. We also adjust Windows and Edge settings — like turning off Bluetooth and location services — to ensure clean, repeatable results.
Yes, yes, the above is all well and good, but if you're into games — who cares? When buying a gaming PC at any price, you want to know what kind of frame rates you can expect and what the playing experience is like from top to bottom. So we amp up everything to find that out.
This is obvious, right? We use a constantly shifting slate of games comprising new and popular older titles that cover the spectrum of available genres, video hardware support, and graphics features.
Our current list includes:
All tests are run at 1920 x 1080p and the system's native resolution, with settings usually at or near the top of the performance scale (on the theory that you want to know the best a computer is capable of doing).
We may also run some newer or more demanding games at lower detail levels, too, to reflect the reality of what most people are likely to experience. In all cases, we report the resulting frame rates; anything at or above 30 frames per second (fps) we deem playable, while 60fps is our threshold for genuinely smooth playability. As with storage space, though, more is almost always better.
The increase in gaming prowess necessitates working a bit more with 3DMark, so we also run that suite's Fire Strike Ultra (4K DX11), Time Spy Extreme (4K DX12), Port Royal (ray tracing), and Speed Way (DX12 Ultimate) tests.
As we said, laptops typically don't get unbearably hot anymore — at least when playing YouTube videos. But scorching frames rates can still be both a figurative and a literal description.
To see how well gaming laptops can take the heat, we do a variant of our YouTube test, except this time running Metro: Exodus. We use the highest (Extreme) setting and run five full loops of the benchmark before doing our temperature readings (at all the same locations as the other heat test) during the sixth. We still see laptops get up to 140 degrees sometimes — don't you want to know that before you put one on your lap?
Dedicated gamers aren't likely to play hardcore 3D games unless their laptop is plugged in. But if you wanted to, how long could you get away with it? UL has a terrific tool for figuring out this critical bit of information in its PCMark 10 benchmark suite.
Under all the same conditions as our Web-surfing battery test, we run PCMark's gaming battery test until the system shuts off. When we turn it on again, we get the time. It's usually between 60 and 90 minutes — not so great. But what did you expect from playing games off-plug?
Apple releases far fewer computers every year than most other companies we work with, so spend a lot less time with them — but we take each new release very seriously, since Apple's passionate customer base certainly does.
We run as many of our standard PC tests on Mac computers as we can, including Geekbench, HandBrake, and Crossmark, plus of course our display tests.
We run whatever games we can (Borderlands 3, Cyberpunk 2077, Shadow of the Tomb Raider, Sid Meier's Civilization VI, and Total War: Warhammer III), though most require some slight tweaks to the settings, resulting in scores that, if not always 100% comparable to what we get on PCs, are as close as we can get.
For MacBook Pro releases and Apple desktops (such the Mac mini, the Mac Studio, and the Mac Pro), we also run the cross-platform Pugetbench for Adobe CC workstation tests.
We also run a different version of our battery test for Apple laptops, but it's based on the same essential idea: visiting websites over and over and until the system runs out of power.
Getting this test to work properly involves some adjustments to the computer's settings, such as disabling color filters, automatic brightness adjustments, and True Tone, and turning off other power-sapping features like iCloud syncing.
Just because Chromebooks are laptops doesn't mean they work the same way Windows PCs do. Sure, some of our same tests still run (Geekbench, heat, display), but most of them are necessarily different because Chrome OS is its own special universe.
Testing Chromebooks is a must, though, as these ultra-affordable systems have become as common as chalkboards and three-ring binders were during my school years (sigh).
And even if you've been out of school as long as I have, these can be handy systems to fill in holes in your personal tech plan — provided you can do everything you need to with Web apps, of course.
Speaking of Web apps! This classic suite combines dozens of JavaScript and Web Assembly tests into a single application that tells you a lot about how well the system is likely to function while Web browsing.
Subtests include cryptography, ray-tracing, image editing, 3D rendering, and loads more. Because we also run this test on phones and Android and iOS tablets, some cross comparison is possible—if you're interested in that sort of thing.
More real world in scope than JetStream 2, WebXPRT also expands on it by utilizing HTML5, which is now common across the Web, in addition to Canvas, JavaScript and Web Assembly.
The application's scenario selection comprises photo enhancement, AI photo album organization, stock option pricing, encryption and OCR scanning, sales graphs, and online homework. The result of the test is a single score that can be compared with that from any other Web-browsing device.
Another Principled Technologies test, CrXPRT 2 has been designed just for Chrome. It runs a set of scenarios similar to WebXPRT 4's (Photo Effects, Face Detection JS, Offline Notes, Stock Portfolio Dashboard, DNA Sequence Analysis, 3D Shapes with WebGL), but focusing on Web browsing, JavaScript, and WebGL, with Chrome OS–optimized programming.
Chrome's move toward greater Android compatibility means we can now run 3DMark on Chromebooks, and we do, so we can better compare their graphics capabilities to those of other lower-end devices. Our exact slate of tests depends on the system's capabilities, but we run everything we can.
Although our standard battery test doesn't work in Chrome OS, we've developed another one that works almost identically.
After using a Chrome plug-in to ensure the screen doesn't turn off, we run our proprietary test to visit a collection of major websites (these live) until the battery runs out.
Workstation PCs have a lot in common with gaming laptops. They have fancier processors, more powerful graphics cards, and tend to be physically bigger — and chances are you didn't drop a ton of money on one just to type up Word documents.
Because everything about workstations is tweaked to optimize certain kinds of performance, gaming tests won't do. So we supplement our everyday performance tests and Civilization VI with some heavier-duty applications.
Puget Systems isn't just a leading manufacturer of workstation computers — it also makes some of the most compelling test suites in the industry, which put powerful computers through grueling challenges drawn from industry-leading programs. Our favorite of these are for Adobe's Creative Cloud Photoshop (photo editing), and Premiere Pro (video editing). We look at both the overall synthetic score and how long it takes to complete each individual test.
Benchmark stalwart UL had to have a say in professional performance, too, and its relatively new Procyon test suite does not disappoint. Its Photo Editing test combines Lightroom Classic and Photoshop into a single series of photo editing and organizational tasks; Video Editing subjects the computer to a particularly punishing Premiere Pro workload.
Like BAPCo, the Standard Performance Evaluation Corporation (SPEC) has been around for ages, and its benchmarks have long been the go-to for determining performance on high-end rigs. We run these tests in part to complement the others, as they are largely graphics driven and even the processor tests dig deeper than we see elsewhere. Workloads in the new, robust 4.0 version include Autodesk Inventory, Blender, LuxCoreRender, Python 3, and more — nearly two dozen in total.
After we have finished running the above tests, the system is then passed on to the reviewer. The amount of 'quality time' the reviewer spends with the computer can range from a few days to a couple of weeks — whatever is necessary to get the best idea of how well the computer functions under everyday conditions.
The criteria applied at this stage are dependent on the system in question, and change slightly from reviewer to reviewer, but usually involves answering questions such as these:
Just as technology and the computer industry itself are always evolving, so too are we constantly re-evaluating and revising our tests to make sure they always provide you the information you need and want.
Check out our How We Test page to see how we evaluate other types of tech gear, from phones to TVs. Have a question or suggestion about our testing procedures? Let us know! Our goal is to help you make the best buying decisions possible, no matter what you're shopping for.
Follow Tom's Guide on Google News to get our up-to-date news, how-tos, and reviews in your feeds. Make sure to click the Follow button.