Running game benchmarks is a good way to test a computer's performance. Modern games are often demanding on the CPU and GPU, pushing their computing capacity to the maximum. The results of these benchmarks not only inform us about the laptop's performance in each game we test, but they also give us a rough idea of how the laptop will perform in other games with similar hardware requirements. More importantly, they allow us to compare different laptops and configurations. This article details how we conduct the game benchmarks and how the results can help your buying decision.
Game benchmarks allow us to gauge a laptop's overall performance because many modern games are extremely demanding on the CPU and GPU. These benchmarks give users data that they can compare between devices so they can make a better buying decision for the type of games they plan on playing. They're also helpful when choosing a laptop for graphically demanding tasks like video editing or 3D modeling because better gaming performance usually translates to better performance elsewhere.
Our game benchmark test setup is fairly simple, as aside from Counter Strike: Global Offensive, all of the games we test have a built-in benchmark tool within the game itself. For Counter Strike: Global Offensive, we use a third-party benchmark tool called CapFrameX (version 1.5.7). We perform the benchmarks in a temperature-controlled room set to 22°C (71.6°F), with a tolerance of ±0.5°C. For Windows laptops, we set the power and performance setting to 'Better Performance', and we ensure that the laptop is fully charged and plugged in with its own power adapter. We run the benchmarks at high settings four times, but we only keep the results of the last three runs because the first one tends to differ from the rest, so we consider it an outlier. We run the benchmark at the lowest graphical settings once.
At this time, none of the games that we test have official support for MacBooks with Apple silicon, so we run all the games through Rosetta 2, which typically results in worse performance. The only Intel MacBook that we've tested is the Apple MacBook Pro 16 (2019), and it'll likely remain the only one for the foreseeable future as all new Apple laptops come equipped with Apple's own silicon. Additionally, the application we use to capture the frame time isn't compatible with macOS, so we can only provide a subjective account of the gaming experience.
As for Chromebooks, none of the games run on Chrome OS. Some have a Linux version, but even then, they only work on Chromebooks that run on an x86 CPU, not an ARM-based one. Also, running games on Linux is a workaround that requires a lot of tinkering, which we don't expect many Chromebook users are willing to do. We don't have any plans to test Android games from the Google Play Store at this time, though generally speaking, most Chromebooks can run them without any issues because they're designed to run on smartphones with even lower-end hardware.
For Borderlands 3, we run the benchmark at the 'High' Overall Quality preset at 1080p in DirectX 12. We choose Full Screen mode because the Windowed or Windowed Borderless modes tend to impact performance negatively. We disable Vertical Sync to ensure that there isn't a frame rate cap.
For Civilization VI, we set the Performance Impact and Memory Impact sliders to 'High'. We set the resolution to 1080p, UI upscaling to 100%, Window Mode to 'Full Screen', and Anti-aliasing to 'MSAA 2X'.
For CS:GO, we run the benchmark at 1080p in Full Screen mode, with the following settings:
We disable Laptop Power Savings.
In Shadow of the Tomb Raider, we use the 'High' graphics Preset, with the resolution set to 1080p, the graphics API set to 'DirectX 12', and Anti-Aliasing set to 'TAA'. We run the benchmark in 'Exclusive Fullscreen' with V-SYNC disabled. To keep a level playing field with other GPUs, we leave NVIDIA's DLSS off.
The average frame rate test is pretty self-explanatory. The goal is to show the average frames per second that the laptop can achieve in each game. While every game is different in terms of optimization, the average frame per second result can give you a rough idea of how a laptop will perform in other games that run on the same graphical engine or have similar hardware requirements. We consider 60fps as 'good' and the minimum for a smooth gaming experience, though some games, like Civilization VI, are playable at lower frame rates because they don't require fast reaction times. The limitation of the average FPS result is that it doesn't show whether there are noticeable stutters or frame rate dips.
The 1% Low test shows the average of the worst 1% frame rate (converted from frame time data) in a particular game. In other words, it gives an idea of how much the game stutters. This testing method has been widely adopted in recent years to replace the conventional way of showing the minimum fps because the minimum fps can be misleading. For example, if a large frame drop only happens once in a game, it isn't representative of the game's overall experience because it only happened once. Averaging out the slowest 1% frame times removes the outlier, giving us a better indication of the gaming experience as a whole. For the 1% Low, you want a number as close to the average fps as possible, as a larger disparity means that the stutter is more noticeable.
The 1% Low test mentioned above is represented in frames per second because it's easy to understand. However, this presents a problem with our scoring because there isn't a 'good' number; the number only matters when we consider the average FPS. This is where the 1% Delta test comes in. It shows the difference in frame time (time between each frame in milliseconds) between the average FPS and the 1% Low, giving us a better idea of how smooth the game feels. For motion to appear fluid, the time between each frame needs to be relatively consistent and free of large spikes. For example, a laptop that can consistently push out 30fps (with the same amount of time between each frame) will feel smoother than one that can push out 60fps on average but with dips into the 20fps range. These dips mean one or several frames need to be held longer, which the user perceives as stutter.
It's normal for the frame time to vary from one frame to the next, but that variability should be minimal, so the lower the number (in milliseconds), the better. A frame time variation of more than 10ms is visually noticeable to most people. We include the frame time graph to make it easier to see frame time spikes. Again, we're looking for consistency, so the line representing the frame time should be as smooth as possible and free of large jumps. Note that the three big frame time spikes in Shadow of the Tomb Raider are scene changes and aren't representative of the laptop's performance.
The 0.1% Low and the 0.1% Low Delta are similar to the 1% Low and 1% Low Delta, but we're only counting the 0.1% slowest frame times. The 0.1% Low fps is roughly the lowest frame rate you'll likely see in the game. These frame rate dips are more noticeable than the 1% Lows and correspond to instances when the game might hang or freeze, rather than just stutter. The idea is the same: you want to look for a higher 0.1% fps and a lower 0.1% delta.
The average FPS on Very Low/Minimum test is the same as the Average FPS test but performed with the lowest graphical settings allowed in each game at 1080p. It's sometimes the only way we can get a game to run on lower-end hardware. The result can be helpful for those looking for a laptop with just enough processing power to run the game smoothly, regardless of the graphical settings. This test doesn't count toward the final score for each game benchmark.
The Average Turn Time is a CPU test and part of Civilization VI's built-in benchmark tool. It measures the time it takes for the computer to work out all the possible moves, so a lower number is better because you spend less time waiting for the AI to make its move. The turn time is generally better on CPUs with more cores and threads, but it's unclear whether there's a limit to the number of cores the game can utilize.
Although we perform these game benchmarks as part of our laptop testing, it's important to understand that every game has different levels of optimization. As such, our benchmark results can only provide a rough estimate of the laptop's overall performance, with the end goal of comparing the data with other laptops. Also, changing the graphical settings in a game can significantly improve performance, so even though we perform all the benchmarks with 'High' presets, you're free to turn some down to get better frame rates. Many online sources can guide you in optimizing the settings, especially for AAA titles known to be graphically demanding.
If you often play games that support DLSS (Deep Learning Super Sampling), NVIDIA's AI upscaling feature, it might be worth considering a laptop with an RTX 20-series or newer GPU. DLSS can dramatically increase the frame rate in compatible games by rendering the image at a lower resolution and subsequently upscale it using AI. AMD has a similar feature called FSR (Fidelity Super Resolution), which also works with some NVIDIA GPUs. However, at the time of writing, most user feedback seems to indicate that FSR is still in its infancy and doesn't provide the same level of image quality as DLSS.
When shopping for a laptop, game benchmarks are a good way to compare performance between devices. They show what kind of performance you can expect when playing certain types of games, and they're also a good indication of how well the laptop will perform for other graphically-related tasks like 3D modeling. For devices that we can't benchmark, like Chromebooks, you can refer to the Basemark GPU test to get an idea of the laptop's relative GPU performance.