(Credit: Joseph Maldonado)
The process of reviewing desktop CPUs at PCMag carries on core traditions that date back to the establishment of PC Labs in 1984: We compare each CPU in its category based on price, features, upgrades from previous generations, and in-house performance tests.
To evaluate performance, we use a suite of software benchmark tests and real-world applications and games, carefully chosen to highlight the strengths and weaknesses of a tested CPU to see how it handles different situations. Our desktop CPU testing breaks down into two rough classes: productivity and graphics.
(Credit: Joseph Maldonado)Modern-day processors are highly complex pieces of technology with multiple distinct components inside. Most nowadays are essentially a system-on-a-chip (SoC) with many of the most critical pieces of circuitry now integrated into the processor. This complicates testing CPUs, requiring multiple benchmarking methods to evaluate each processor’s capabilities. Here, I’ll review how we benchmark processors at PC Labs and the test systems I use.
The Test Hardware
As AMD and Intel CPUs are mutually incompatible at the motherboard level, I maintain a separate test bed for each brand of chip. I aim to maintain parity in all sensible areas. Both of my late-model test beds for consumer CPUs (for AMD AM5 and Intel Socket 1851) are Praxis Wetbench open-air cases with separate but identical SilverStone Hela 1650R Platinum (1,650-watt) power supplies and Corsair iCue Link Titan 360 RX LCD CPU coolers. Each system also houses a 1TB PCIe 4.0 M.2 SSD with Windows 11 installed. I also maintain other testbeds for legacy sockets like AMD AM4 and Intel LGA 1200, and build out new testbeds as needed for special releases like AMD's Ryzen Threadripper as those chips release.
(Credit: Joseph Maldonado)The memory capacity is identical on both main (AM5 and LGA 1851) systems, with two 16GB sticks of DDR5 RAM, but the same RAM is not always used in each system. One of the processor’s many parts is the memory controller, which typically improves over time. I test all processors using RAM clocked at their max official clock speed so that improvements in this area contribute to any change in performance. An Nvidia GeForce RTX 5090 graphics card is the only internal component swapped between the two systems.
(Credit: Joseph Maldonado)I currently use a Gigabyte X870E Aorus Master motherboard for AMD-based socket AM5 processors. I test Intel processors supporting LGA 1851 using an ASRock Z890 Taichi motherboard.
Cinebench 2024
Maxon has been crafting new versions of its Cinebench software for more than a decade now, and the benchmarking app is now a clear industry favorite as a CPU-performance drag race. The Cinebench benchmarks are built to use the same rendering engines as Maxon’s multimedia content creation tools, and they help to show how well a PC will perform when rendering content made with that software.
(Credit: Maxon)The latest version of Cinebench I use is Cinebench 2024, based on Maxon’s Redshift rendering engine, which is employed in the company’s Cinema 4D software. It renders a scene continuously for 10 minutes and then finishes rendering the final scene when the timer runs out. The Cinebench score we report is based on how quickly, and how many times, the scene is rendered.
We run this test in two iterations: to measure multi-threaded performance and single-threaded performance separately. The score reported in each case is meaningful only in relation to other Cinebench 2024 scores.
Blender Benchmark 3.1
Blender is another popular tool for creating digital content like 3D models and 3D environments. Blender's developer has recently released the Blender Benchmark software as a standalone testing utility (which works with various versions of the Blender core program) that helps to provide standardized and consistent benchmark results. We currently use Blender Benchmark 3.1 and, within it, select Blender version 4.3.0 to run for all tests.
(Credit: Blender)This test tends to respond strongly to increases in core count and clock speed, but shows less change from different graphics cards installed across the same CPU. It renders three sample models called Monster, Classroom, and Junkshop, kicking back a score in samples per minute. Those are the scores we report, useful in comparison to other tested CPUs.
Power and Thermal Testing
During the Cinebench and Blender tests, I also measure the benchmark system's power consumption and the CPU's operating temperature. Both tests are intensely demanding on the hardware, pushing the parts toward 100% utilization. Cinebench 2024, with its 10-minute persistent workload, helps to show what power consumption and operating temperatures look like when running a complicated task for a long time.
(Credit: CPUID)The Blender Benchmark is different in this regard from Cinebench 2024. Though the workload is still quite strenuous on the hardware, the Blender Benchmark runs quite quickly—in just a couple of minutes. (The actual test running time depends on how fast the tested CPU is at completing the test, but it usually doesn’t take long.) In between the three scenes rendered by the Blender Benchmark, the system will ramp down for a moment as the data for the new test is loaded.
Checking power and thermals while running the Blender Benchmark gives us insight into how power and thermals respond to sudden spikes in workloads. The system is idle before the test and doesn’t always get as hot as during the Cinebench test, as it only needs to work hard for a shorter period.
I use a Kill-A-Watt wall meter to measure power consumption during these tests. I typically use the CPUID HWMonitor software to measure the operating temperature. Usually, these work, but on rare occasions, I have encountered times when HWMonitor doesn’t display operating temps for a component in testing. In this case, I usually switch to Open Hardware Monitor. These programs are similar, and while I have found CPUID’s HWMonitor to generally have better compatibility, Open Hardware Monitor sometimes works to show operating temperatures for components that HWMonitor doesn’t.
HandBrake 1.8.2
Converting videos from one format to another is often slow unless you have powerful hardware. I’ve used this test for some time now and have had some processors take more than 40 minutes to transcode the short 4K clip of a publicly accessible short film known as Tears of Steel I use to 1080p, but most mainstream and high-end chips these days typically take just a few minutes. I now run this test using HandBrake 1.8.2, the latest version available when I set up our current CPU test beds.
This test appears to favor higher core counts over all else, though increases in clock speed also positively affect performance. The conversion time is the score reported; lower times are better.
(Credit: HandBrake)POV-Ray 3.7
POV-Ray has been around for ages and is another industry favorite for testing CPU performance. As the software’s name suggests, POV-Ray creates highly realistic images using ray-traced lighting and illumination. The software has a built-in benchmark that can run in a single-threaded or multi-threaded mode, enabling us to gauge the performance of both. Today, the fastest CPUs often complete the multi-threaded test in less than 30 seconds, while the single-threaded test takes several minutes. Like with Handbrake, the completion time is our reported score, and lower times are better.
(Credit: The POV-Team)Adobe Photoshop and Premiere Pro
Adobe is a company that scarcely needs an introduction. Its software has long been synonymous with content creation, and we test its Photoshop and Premiere Pro applications to see what sort of content creation performance you can expect from the CPUs we test.
Photoshop noticeably responds to increased CPU performance, while Premiere Pro spikes in speed as both CPU and GPU performance rise. As we use an Nvidia GeForce RTX 5090 for all graphics cards we test, however, any change in test scores should be down to CPU differences alone.
(Credit: Puget Bench)We use a utility from the workstation maker Puget Systems to test both programs; it is called PugetBench for Creators. PugetBench runs both Adobe applications through a standard set of tasks and typical creative workflows. Photoshop is run through file resizes, image rotation, smudging, layer manipulation, and lots of filter applications. Premiere Pro gets subjected to a host of encoding, decoding, and processing tasks with a variety of codecs and file types. The PugetBench score reported is useful only in relation to other PugetBench scores; higher is better.
3DMark Time Spy Extreme
Gaming performance is critical to many looking to buy a PC, and the interplay between your CPU and the graphics card you own is complex. (In some cases, an underpowered CPU can hold a GPU back from reaching its full potential.)
UL's 3DMark is another industry favorite tool for benchmarking gaming performance. You’ll find several tests available in 3DMark, but I focus on the Time Spy Extreme test, as it is the most recent 3DMark test that looks explicitly at CPU performance in addition to overall gaming performance.
(Credit: 3DMark)The test results from 3DMark do not directly reflect real-world performance in games, as it is a synthetic test, but the results indicate relative performance for actual games. The scores are proprietary to 3DMark and only useful when comparing one Time Spy Extreme score to another. Higher is better.
Cyberpunk 2077
The AAA game Cyberpunk 2077 is regularly updated and has become one of the industry’s most heavily used gaming benchmarks due to its support for a wide range of new gaming technologies. The game is also highly demanding.
I benchmark processors via this game using two different graphics presets. I test it first with the Ray-Tracing: Ultra preset, just one step below the highest graphics preset. The only setting I change in the standard preset profile is to disable DLSS, which shows how well the CPU will perform under these conditions. In this test, the graphics card is the bottleneck when paired with most high-end CPUs, but slower CPUs will still return lower performance under these conditions. It’s also important to see how much of a gain, if any, you get from one CPU over the other when running with high settings like this, which not everyone considers.
(Credit: CD Projekt Red)After running the test with the Ray-Tracing: Ultra preset, I benchmark the processor again using the low preset. Here, the graphics card is much less of a bottleneck, and this tends to show more divergence among CPUs.
We test using both presets at 1,920 by 1,080 (1080p or full HD), 2,560 by 1,440 (1440p or QHD), and 3,840 by 2,160 (2160p or 4K) resolutions. We report frame rates for this and all subsequent games below; higher numbers are better.
F1 2024
I run F1 2024 in a similar way to Cyberpunk 2077: I first run the test using the Ultra High graphics preset, and then rerun it with the Ultra Low preset. I benchmark this way for the same reasons listed for Cyberpunk 2077. I also test CPUs via this game at all three previously mentioned resolutions using both presets.
(Credit: EA Games)Black Myth Wukong
With Black Myth Wukong, you are forced to use some form of screen optimization technology, such as AMD’s FSR, Intel’s XeSS, or Nvidia’s DLSS. The CPU test bed I use is equipped with an Nvidia GeForce RTX 5090, which is always set to use DLSS, but frame generation is kept off. I test this game at all three resolutions using the cinematic and low graphics presets.
(Credit: Game Science)Total War: Three Kingdoms
Total War: Three Kingdoms was released in 2018 and isn’t quite as graphically demanding as the newer games I use to test. It’s also a real-time strategy (RTS) game that generates thousands of models on screen during a battle, which makes it an excellent test of a CPU’s gaming prowess.
(Credit: Sega)I only test using this game at the medium preset and at resolutions of 1,280 by 720 (720p) and 1080p. After testing this game with the Nvidia GeForce RTX 5090, I also use this game to test the CPU's integrated graphics processor (IGP) using the same graphics settings.
Shadow of the Tomb Raider
Shadow of the Tomb Raider was also released in 2018, and it remains a popular benchmark due to the game’s still-impressive graphics. I test this game using the low graphics preset and with TAA antialiasing, while the resolution is set to 720p and 1080p. After that, I also retest the game using the processor's IGP to measure that component's performance.
(Credit: Square Enix)An Overall View of Processor Performance
With the data I gather from these tests, we can gain insight into a processor's performance in several vital areas, including content creation, media transcoding, and gaming. We can also understand a processor’s energy efficiency and heat generation.
(Credit: Joseph Maldonado)IGP performance isn’t that important to most people, but the test data I gather also tells us something about this. As someone who has used IGPs to play games enjoyably in the past, this is also an area of particular interest to me. Most people will be more interested in seeing the performance results from tests with the GeForce RTX 5090, as this can show you how well processors perform in games against other chips, and whether a given CPU might be a bottleneck in games versus other possible CPU choices.
The Result? Authoritative Guides to Buying the Best CPUs
This detailed benchmarking process is critical to how PC Labs informs the content and choices within our roundups and reviews on PCMag.com. Performance is just one factor in the overall picture we take of a processor’s worth. We also assess design, feature set, and value for money.
(Credit: Joseph Maldonado)Using the data from our benchmarking process, we maintain three different processor buying guides: the best CPUs overall, the best CPUs for gaming, and the best budget CPUs. We update these guides in sync with major market releases and, especially, when we bestow our Editors' Choice award on a given product.