Benchmarking is one of the most popular pastimes of the PC community, whether the goal is to simply measure the performance of a PC or to get bragging rights on leaderboards. It’s not difficult to run a benchmark; you just have to download some software, run it, and record the results. From casual users to professional reviewers, benchmarks are practically universal.
But there is a bit more to benchmarking than just running a test; there are a few key things to consider. To be clear, there isn’t a single correct approach to benchmarking individual components, and in fact, there are different methods of benchmarking that can get useful data. Good benchmarking is really about understanding what you’re testing, under what conditions, and how you’re testing it.
Related
5 reasons I love benchmarking
Ever since I got into PC building and gaming, benchmarking has been one of my favorite hobbies, and here are five reasons why.
Two kinds of benchmarking
Scientific vs. realistic benchmarking
The first question you ought to ask yourself is why you’re benchmarking — you either want to find the maximum theoretical performance of a component or how it performs in typical day-to-day use. The former approach is scientific benchmarking and the latter is realistic or practical benchmarking, and this is generally where the benchmarking practices of a reviewer and a user will diverge.
In many CPU reviews, you’ll see the test system using really low graphics settings no matter what CPU is being tested. This is to ensure that the CPU, and not the GPU, becomes the limiting factor (referred to as a bottleneck) in the performance being measured.
The method of benchmarking a component scientifically is about isolating all other components to make sure they’re not influencing the results of a benchmark. For instance, in many CPU reviews, you’ll see the test system using really low graphics settings no matter what CPU is being tested. This is to ensure that the CPU, and not the GPU, becomes the limiting factor (referred to as a bottleneck) in the performance being measured.
Other software-related practices include testing on a freshly installed Windows system, closing as many background tasks as possible, and setting the power settings to maximum. Basically, you want every component of a test system (except the one that’s being tested) to be as high-end and clean as possible, so that it doesn’t factor into the data.
Obviously, none of us use our computers like this, and the results from scientific benchmarks would be essentially unrealistic, which is why you’ll probably use realistic benchmarking with your rig (unless you’re a reviewer). For end users, this means running benchmarks under normal usage conditions, with the usual background applications running, and with regular settings and regular hardware.
If you’re only trying to get a quick score for your hardware, or trying to optimize the performance of your PC, realistic benchmarking is probably enough for you.
Reviewers might sometimes take a pseudo-realistic approach for scientific benchmarking, where the conditions are still controlled, but tweaked to reflect a typical user’s PC.
If you’re trying to hit the highest benchmark scores for bragging rights or pure curiosity, scientific benchmarking is probably what you want to do. But if you’re only trying to get a quick score for your hardware, or trying to optimize the performance of your PC, realistic benchmarking is probably enough for you.
Before you benchmark
Download essential monitoring tools
Benchmarking isn’t always an automatic process. Depending on the benchmark and what you’re testing, you’ll need additional applications to record and monitor important parameters for your components. For the most basic hardware monitoring, Task Manager is perhaps the most well-known tool. It’s commonly used for killing programs, but it also comes with a decently robust hardware monitoring section for the CPU, RAM, storage, graphics, and network devices.
Task Manager only shows diagnostic data and can’t record it, but if you’re just curious about basic information like component usage, then it should be sufficient.
HWiNFO is a great program for showing this type of data in a simple tabular format, as it not only tracks real-time data, but also minimums, maximums, and average values.
In most cases, however, you’ll need slightly more in-depth information about frequencies, power usage, temperatures, voltage, and other data points. HWiNFO is a great program for showing this type of data in a simple tabular format, as it not only tracks real-time data, but also minimums, maximums, and average values.
There’s also MSI Afterburner, which tracks much of the same data but shows it in a graph form or even in an overlay in games, which is useful for viewing critical component info while running a gaming benchmark.
HWiNFO
HWiNFO is a lightweight and simple hardware monitoring tool that lets you check CPU and GPU temperature, power draw, frequency, voltage, and more.
MSI Afterburner
MSI Afterburner is one of the best overclocking software with many extra features to monitor your PC’s performance.
If you plan on benchmarking games without using built-in tests, then you’ll need a program that records framerates, and one of the best applications for that is OCAT. It’s an open-source program that records the average and 99th percentile framerates, visualizing them on a graph. OCAT also comes with a useful overlay that you can disable if it causes problems (which isn’t infrequent).
The best way to test a game with OCAT is to create a save, start recording, and then play the game normally for a few minutes. If you want to run multiple, repeatable tests, just reload the save, and play the game normally again for the same amount of time.
OCAT
OCAT is an open-source capture tool that you can use to record important CPU and GPU parameters during games and other applications.
If you’re doing casual benchmarking, it’s probably not necessary to use any of these applications. But for some things like power consumption and voltage, you pretty much have to use an app like HWiNFO or MSI Afterburner, as benchmarking software rarely ever records that info itself.
Know your CPU’s limits
Okay, we’re finally going to talk about individual components and how to test them. The CPU is one of the most commonly benchmarked PC components. Since it is important for both gaming and non-gaming workloads, knowing if your CPU is performing to its fullest potential is often a top priority for most users.
If you’ve bought one of the best CPUs on the market, you’ll probably want to benchmark it to ensure you’re extracting the maximum performance. Many CPU benchmarking tools suit different needs and use cases — including popular free programs and paid tools with in-depth benchmarking features.
Cinebench 2024
There are three basic things you can test with CPUs: single-core performance, multi-core performance, and gaming performance. These three performance areas have some overlap but are pretty distinct. Cinebench 2024 is a great free benchmarking tool for testing raw single- and multi-core performance. It has evolved a lot over time and is one of the most well-regarded tools for taxing your CPU.
Cinebench 2024 uses Redshift, Cinema 4D’s rendering engine, to measure your processor’s single- and multi-core capabilities. It allows you to run a single pass for some quick testing, or longer tests if you want info on sustained results. You can also compare your CPU’s performance against the scores recorded for the exact model to understand if something is holding your processor back.
Cinebench 2024
This new version of Cinebench adds GPU testing to its excellent CPU benchmarking tests, making it a good fit for those looking to download a simple and easy-to-use benchmarking tool that mimics real-world usage.
PCMark 10
PCMark 10 has been around for a long time, thanks to its collection of tests that mimic real-life CPU usage. It runs office apps, web browsing, video streaming, and gaming tasks to measure how your CPU handles different workloads. PCMark 10 will also tell you the kind of CPU usage and temperatures you can expect during those workloads.
While it has a free version that’s great for most users, you can unlock all the CPU tests along with extra features for $30. One of its best features is the ability to upload your benchmarks to its online database, and compare how your hardware fares against that of fellow users.
PCMark 10
PCMark 10 is a great tool to use if you want to benchmark your PC and see how it performs under load. It runs a series of tests to help you understand your CPU’s capabilities and stability when it’s under load.
Prime95 and others
When it comes to stress testing your CPU, there’s nothing better than Prime95. Stress testing involves subjecting your CPU to extreme loads and seeing how far you can push it before it throttles. It’s not indicative of real-world usage, but it can still come in handy when you want to be absolutely sure that your overclocked processor will be stable under the heaviest workloads.
Blender Benchmark is a more realistic benchmark for testing your CPU’s multi-threaded prowess. It runs a short CPU test to simulate real-world workloads and also allows you to upload the results online. Lastly, PugetBench is good for showcasing general performance in creative applications like Adobe Photoshop and Da Vince Resolve.
Prime95
Prime95 is a stress-testing tool that puts a tremendous load on the CPU to test its performance and stability. You can download and use it for free, but remember that it’s recommended for beginners.
Gaming benchmarks
Testing CPU performance in games is quite complicated, but we’re going to keep it simple and focus on some popular titles for testing CPU performance. There are tons of games out there, but I’d recommend testing three types of games: a high FPS esports title, a modern, graphically intensive AAA game, and an older AAA game.
Great games to test include Counter-Strike: Global Offensive, Cyberpunk 2077, and Grand Theft Auto V for each respective category. Simply testing games that a lot of reviewers test would also be a good idea.
Can it run Alan Wake 2?
With the best GPUs, we mainly care about performance in games and creative applications. While most of the GPU benchmarking tools focus on measuring the graphics card’s gaming capabilities, there are some that you can rely on for measuring video editing and 3D rendering performance. If you’re overclocking your GPU, you’ll probably need at least one of these tools to test your GPU’s stability before taking it for a spin during an actual gaming session.
Even if your overclock seems stable during a stress test, it still pays to test it during a real gaming scenario to ensure 100% stability.
3DMark
Source: Steam
3DMark benchmarks like Time Spy and Firestrike have been around for decades, thanks to their simple and easy-to-use methodologies for testing a GPU, and getting an easy-to-understand score. Essentially, you choose a few settings, pick the specific benchmark you need to run, and then wait for the final score.
It might not fully emulate realistic gaming performance, but it’s one of the most popular GPU benchmarks in the world.
The tests use your graphics card to render various graphically demanding scenes and return a composite score along with graphs for frame rate, temperature, frequency, and GPU load.
It might not fully emulate realistic gaming performance, but it’s one of the most popular GPU benchmarks in the world, so you can easily compare your system’s performance against that of fellow PC users.
3DMark
3DMark is a popular benchmarking utility that includes a collection of benchmarks to test your GPU’s performance with various settings. The free version of the program is enough to run some basic tests, while the paid version unlocks additional tests to collect more information.
Superposition
Superposition is another popular (and free) GPU benchmark that’s extremely easy to use. The free version allows you to choose a resolution and quality setting before running the test. It returns a composite score along with the average, minimum, and maximum FPS achieved during the test.
You can also see GPU temperature and clock speeds in real time as the test is running. The paid version provides looped tests, leaderboard posting, stress tests, and more, but it’s not worth it for most casual users.
Superposition
The free version of Superposition gets you a basic benchmarking tool that’s enough to gauge your GPU’s performance and stability. It even includes a free VR benchmark tool, which is great.
FurMark and more
Source: Geeks3D
FurMark is an extremely demanding stress test for your GPU for when you feel like pushing it to its limits and beyond. Its various tests use your GPU to render different kinds of 3D fur, providing useful information about your GPU such as temperature, clock speed, and fan speed.
FurMark is an unrealistic test and should only be used if you aren’t satisfied with other GPU benchmarks.
Some people find FurMark to be too demanding and dangerous for your GPU, and there’s good reason to feel that way. It’s an unrealistic test and should only be used if you aren’t satisfied with other GPU benchmarks.
Cinebench 2024 now has a dedicated GPU benchmark, if you want a tool to test both your CPU and GPU. For specifically testing your GPU’s mettle in creative workloads, Blender Benchmark is a good test because it runs a realistic scenario that most people use in their day-to-day workloads. V-Ray is another rendering benchmark that can be used to test the ray tracing performance of your Nvidia GPU.
Also, PugetBench is great for testing performance in Adobe applications, Da Vince Resolve, and Lightroom Classic. But keep in mind that your CPU can also have a significant impact on these results.
FurMark
FurMark is also a popular benchmarking program, although it’s more suitable for enthusiasts who often push their GPUs to their absolute limits and know what they are doing.
Gaming benchmarks
Red Dead Redemption 2, Quality Thermal Paste
Of course, benchmarking your GPU with the best games is often the easiest and most realistic benchmark. You can just run a fairly modern game with your desired settings and see how fast your GPU performs. You might also want to try messing around with different settings like ray tracing, DLSS/FSR/XeSS, DX11 vs. DX12, different resolutions, and so on to get a complete picture.
You can use MSI Afterburner to display an in-game overlay for FPS, temperature, and GPU load monitoring.
Gotta benchmark that fast DDR5 RAM
Although there are some RAM benchmarks like the one included in AIDA64 that show you the read/write speeds and latency of your RAM, it doesn’t tell you much about the real-world performance of your PC. Instead, you should run CPU benchmarks (primarily games), which are most likely to benefit from, say, faster DDR5 RAM, and see if there’s any performance uplift.
AIDA64
AIDA64 is a system monitoring and benchmarking tool that offers detailed information, diagnostics, and stress tests to measure the performance of your computer.
You’re only as fast as your SSD
For most users, CrystalDiskMark is the only benchmark needed for testing the best SSDs. It tests pretty much all the main metrics you’d care about in an SSD or HDD (sequential and random workloads), and is also highly customizable. If you want a benchmark with a score attached to it, then 3DMark Storage Benchmark is a good choice.
CrystalDiskMark
CrystalDiskMark is one of the most popular storage benchmarking tools. This open-source program tests the speed and efficiency of your HDD or SSD by writing data in sequential and random tests.
Benchmark your CPU and GPU cooler
Cooling can be your Achilles heel
Testing the cooling prowess of a CPU cooler hinges on creating as much heat as possible by letting your CPU consume as much power as possible. Cinebench 2024 comes in handy when you need to test your CPU cooler while monitoring your CPU temperature with HWiNFO. For GPUs, you can either run Cinebench 2024 or 3DMark Time Spy, and monitor the GPU temperature with HWiNFO or MSI Afterburner.
Once you observe the behavior of your CPU and GPU cooler under load, you can use Fan Control to set fan curves for both. You can balance fan speeds and noise levels for the most optimum performance.
Benchmarking is as difficult or challenging as you want it to be
Whether you’re a casual user just looking to quantify your PC’s performance, or a professional trying to figure out the limits of a component, benchmarking can be useful and even fun. One of the greatest things about benchmarking is that you get to determine how far you take it, since there’s no single right way to test hardware.
To be clear, some ways are better than others, but as long as you’re consistent in how you benchmark, you should be able to get usable data, which is the whole point of benchmarking.