To fully understand your PC’s performance, you need to understand benchmarking and benchmarking software. This tutorial helps you do that by walking you through all you need to know to benchmark your PC and the benchmarks to start with.
What Are Benchmarks?
In this context, a benchmark is a fixed test (measure) of performance – whether of your entire PC or individual components. This measure of performance can be lined up to and compared with other PCs that have run the same benchmark, allowing you to see how your performance lines up compared to everyone else’s.
Is one of your components underperforming in comparison to what other people are scoring? A benchmark may reveal that discrepancy.
Have you overclocked your components? A benchmark will help you quantify how much improvement you made and test the system’s stability.
How do I benchmark my PC?
Fortunately, benchmarking software is usually pretty simple, especially if you aren’t specifically running a graphics-focused benchmark. As a quick example, below is the installation and initial process of Geekbench 5.
1. Download Geekbench 5 to your Downloads folder of choice. Once the download finishes, run the setup application.
2. From here, it’s just your usual Windows installation process – reading the EULA, deciding if you want it pinned, and so on. You can even launch it right out of Setup, and after choosing the trial run, your benchmarks are immediately available.
3. Click on the “Run Compute Benchmark” button to start the benchmarking test.
This straightforward download-to-setup-to-run process is pretty much identical across top benchmarking software. With the benchmarking software that I test in this article, I’m specifying where the installation process differs or more options are available.
Where more options are available in a piece of benchmarking software, it’s usually graphics options or other benchmarks. You’ll be walked through these options, what they mean, and if you should even use them for all of the software tested below.
The Best Benchmarking Software
Note: I have actually installed, run, and included my results in all of these benchmarks below! My PC’s specs for running these benchmarks are:
- Motherboard: Gigabyte GA-Z490i AORUS Ultra
- CPU: Intel Core i7-10700K
- CPU Cooler: Noctua NH-L9i chromax Low Profile CPU Cooler
- RAM: 16 GB DDR4-3200 Corsair Vengeance RGB PRO (XMP disabled, so clocked at 2133 MHz)
- GPU: Gigabyte Windforce Nvidia GTX 1070
- SSD 1 (boot): SATA3 ADATA ULTIMATE SU800 512GB
- SSD 2 (storage): NVMe XPG Gammix S11 Pro 1TB
1. Geekbench 5
Geekbench 5 is a cross-platform benchmarking tool. We’ve talked through the Windows setup process for this one above, but the specifics are what you should expect: download, run the installer, and run the program. Once you’ve done all of this, you’ll be presented with your choice of two benchmarks.
With “CPU Benchmark,” you can test the speed of your CPU, both per-core and with all cores working together.
With “Compute Benchmark,” you may have a choice of different APIs to use. I recommend using Vulkan, if available, as it is the most modern standard. The “Compute” refers to GPU Compute and is used to benchmark GPUs rather than CPUs.
On an Nvidia GPU, you may also see CUDA as an option under “Compute Benchmark” – this will score lower than the other tests for the reason that it’s focused specifically on a part of your GPU that normally isn’t used by other applications, even games. CUDA is mainly for accelerating professional workloads, like rendering video, without hurting GPU performance.
Once you have your results, you can compare them with any other set of results on the site.
To compare, I’ve also embedded below the results of the same test run off my phone. (To get it for your own phone, check the Geekbench site’s downloads page.)
According to this benchmark, it takes every CPU core of my phone to compare to even one of my desktop CPU cores in raw power. To compare results between your own devices, run the benchmark for yourself and use this page to find other results.
2. Unigine Valley
Unigine Valley is not the latest Unigine benchmark, but it is the best one for running stability tests on a wide range of graphics hardware, including low- and mid-range hardware.
After a quick download and install process, type “Valley Benchmark” in Start and open the benchmarking suite. As with all benchmarking software, you’ll want to use a standard preset to properly compare to other users and hardware. I recommend “Extreme HD Preset” (even for modern low-end GPUs, due to the age of this benchmark). I’ve included my results with that preset at the bottom of this review.
If you’re looking to push the limits of your benchmark even further than this standard and have a compatible display (or graphics card that supports downsampling), you can push the rendering resolution past 1080p by switching Preset to Custom.
Now you get to choose each individual setting – including resolutions besides 1080p. 1440p and 4K, for instance.
In this case, I would recommend turning down anti-aliasing, since it isn’t as noticeable when you’re downsampling or running higher-than-HD signals at native res. This will keep the benchmark in line with what you can expect running most modern games at higher resolutions on your hardware.
“Quality” will have the biggest impact on specific graphical effects and fidelity for each of its tiers. I recommend keeping it on “Ultra” to account for the age of the benchmark, but you can definitely try to lower it if you’re interested in seeing how your hardware scales.
“Stereo 3D” is for 3D monitors. If you don’t have one, ignore that setting.
Once you’ve selected or customized your preset, launch into the benchmark by clicking Run.
At the top of the screen, you’ll notice a row of in-Benchmark options.
Clicking “Benchmark” will immediately start the benchmarking run, which will rotate through various high-fidelity scenes, testing different graphical features while recording system metrics. Your final score will be given in a results screen like the one at the end of this review.
“Camera” allows you to adjust field of view and camera settings, including complete free camera control and a first-person walking camera. The Valley is quite large if you choose to explore it, so be mindful of that.
In-Benchmark Settings and Quality Settings are about the same – mostly. Within Settings, you’ll find a few new options for toggling the FPS/GPU monitor and heavier graphics settings like Volumetric Shadows. I recommend keeping all of these enabled, especially GPU Monitor (to see if you’re overheating under high load).
That’s everything important explained. Now for the results on my system below running the recommended Extreme HD Preset. I’ll explain the score, too.
The FPS score is what is going to matter most here. For the majority of modern games, you’ll want to aim for a score of 60 FPS or better. This means your GPU should be in roughly the punching weight needed for modern games at high settings: 1080p and 60 FPS.
My results here show an average of 95 FPS for my GTX 1070.
According to these GTX 1070 July 2021 benchmarks, the GTX 1070 on High settings at 1080p can push about 88 FPS on average in the latest games. That is uncannily close for a synthetic benchmark, especially one this old, but it shows just how scalable both it and GTX 1070 still are.
At the default 1080p, though, you will be limiting the ability of high-end GPUs to use their faster and larger-density memory. If you want to really see what a modern high-end graphics card is capable of and aren’t just testing stability and temperatures, scroll down to the 3DMark review!
UserBenchmark is among my favorite full PC benchmarking tools for a multitude of reasons. First and foremost, using the application is exceedingly simple: it doesn’t need to be installed, and after you launch the application, you can benchmark your entire PC in one click. Once the benchmark finishes, you’ll see your results automatically opened in your default web browser on a page like the one in the screenshot below.
However, I would be remiss not to discuss UserBenchmark’s weighting of results. While UserBenchmark’s individual metrics will be generally accurate, the final score used to tally up everything can be less consistent, especially if you’re comparing AMD and Intel CPUs.
UserBenchmark remains useful due to the sheer sample size of its database, and it serves as an especially good way to make sure your hardware is performing as it should compared to identicalhardware. However, it’s important to NOT use it to make final hardware buying decisions, since the UserBenchmark-weighted score and real-world comparisons can differ.
3DMark is one of the premier graphical benchmarking solutions on the market, and like Geekbench, is also multi-platform. Unlike Geekbench, it has a variety of paid demos and tiers and essentially has the most intensive CPU and GPU benchmarks on the market at a given time. This is the benchmark to use when you really want to put gaming-grade hardware to the test.
For PC users, especially those who wish to test it for free, setup requires a few extra steps. Specifically, you’ll need to go to the 3DMark Steam Page and click “Download Demo” to get the free trial of 3DMark. From there, you would launch it like any other Steam game.
However, 3DMark is a little different from the other benchmarks on this list, since it’s actually a benchmark suite rather than a singular benchmark on its own. Let’s take a look at the main launcher together.
By default, you’ll see whichever of the three benchmarks included in the demo that 3DMark decides is best for your system.
The current-gen 3DMark test included with the demo is “Time Spy,” so I recommend running that one if you want to see whether your current-gen gaming PC is up to snuff.
However, there are tests besides Time Spy! Let’s take a moment to peek at the others. Click “Benchmarks” in the top-right taskbar.
Some of these Benchmarks will be grayed out due to paid tiers, but they all include explanations for what kind of hardware they’re made for.
The free tests for all users are Time Spy for DirectX 12, Fire Strike for DirectX 11, and Night Raid for DirectX 12 integrated graphics users. Time Spy and Fire Strike are both viable for modern desktops, but Night Raid should be scalable to the integrated graphics in your laptop as well, so long as it was made after mid-2015.
Beside choosing your benchmark, there isn’t much to customize in the demo version of 3DMark. The actual settings are best left on default if you don’t know what you’re doing, and you can’t actually take control mid-benchmark as with Unigine. You’ll need to wait it out to get scored or cancel it entirely by hitting ESC.
Since Time Spy is the most modern test, and you’re unlikely to be using 3DMark without gaming-grade hardware, I’ve chosen to run Time Spy and include my results below.
You’ll get a few different scores. The main two scores are Graphics and CPU score, which measure your GPU and CPU power, respectively. “Estimated game performance” is a metric that is locked behind paid tiers but allows you to see lab-tested estimates for your performance in specific games.
Your main Score is a little different. The green meter/worded score and the numeric score don’t actually match each other.
Your numeric score is for your ranking in the benchmark as a whole. An RTX 3070 gets a score of 12311 compared to my 1070’s 6377, for instance. Lined up with the 1070’s real-world performance of ~88 FPS average in modern games at 1080p, I’d say you’re doing very well once you breach 6000.
The green meter/worded score actually corresponds to how you performed compared to people with similar or identical hardware – below “Good,” and you’re most likely overheating or being bogged down by background programs.
Frequently Asked Questions
1. What should I do before I benchmark my PC?
Before you start benchmarking your PC, it’s important to ensure that you have all other programs closed. Having other applications open will bog down your results, sometimes significantly, especially in the case of games and web browsers. You’re also advised to close as many background applications as you can, but those shouldn’t matter too much.
2. Should I benchmark my PC?
If you’re an overclocker: absolutely, you need to test the stability of that overclock.
If you’re a regular user: It depends. If you’re here, chances are you’re at least curious. We recommend anyone run a benchmark if they suspect that their PC is underperforming. Despite some controversies (discussed in its entry), UserBenchmark is great for identifying whether components in your system are underperforming compared to identical components in other systems.
Outside of testing stability, there’s no reason you have to benchmark your PC. But it can tell you a lot about what you’re working with, especially if you’re looking to compare your current performance to potential upgrades.
3. Should I benchmark my PC before selling it?
Absolutely, especially if you want to quickly demonstrate the legitimacy of your listing and asking price. If you’re selling a gaming PC in particular, you may even want to run games with in-game benchmarks and include those results with your listing.
4. My PC shut off during a benchmark, what do I do?
If you’re an overclocker: congratulations, you just experienced your first unstable overclock! Turn things down and try again or revert to stock and stay there.
If you’re a regular user, chances are high that something is wrong with your PC, most likely your cooling or power. Use a hardware monitor to verify that your components aren’t overheating when you’re running your benchmarks. If you are overheating, it may be time to replace your cooler or thermal paste. (You can also take the steps outlined in this article to alleviate the issue in the meantime.)
If your thermals don’t seem to be the problem, it’s most likely a case of your GPU trying to draw more power than your PSU can handle.
5. What do I do if my hardware is performing below expectations?
Simply take whatever steps you need to to fix the issue. Benchmarking software with a large database is ideal for finding issues like this.