How to Read and Understand CPU Benchmarks

There are several ways to judge the relative performance of CPUs. Comparing their listed specifications is a good place to start: A CPU with a higher core count and faster clock speed will generally outperform its siblings in the same product generation.

But what if you want to compare a top-of-the-line CPU from a few years ago to a midrange one that came out this year? Or what if you need to estimate the speed gains a new CPU can provide for a specific game or application?

That’s where CPU benchmarks come in.

What Are CPU Benchmarks?

Benchmarks allow for easy comparison between multiple CPUs by scoring their performance on a standardized series of tests, and they are useful in many instances:

• When buying or building a new PC. Use benchmark scores to gauge a system’s ability to run games and applications before making a purchase. Lists of scores can be found on review sites like Tom’s Hardware.
• Before upgrading the CPU. Compare benchmarks for various CPUs on a review site — or better yet, on a benchmarking software company’s website — whenever you’re planning a CPU upgrade.
• After upgrading the CPU. Run your own benchmarks to get a preview of the user experience after a major component upgrade.
• Overclocking the CPU. Benchmark your CPU using a program like Intel® Extreme Tuning Utility (Intel® XTU) to evaluate performance gains after an overclock.

CPU Benchmarking Tests & Tools

CPU reviewers rely on a range of different CPU benchmark tests to evaluate CPUs. These fall into two categories: Synthetic and real-world.

Synthetic benchmarks

Use synthetic benchmarks when looking for a quick, general comparison between CPUs.

Synthetic tests simulate many different tasks: 3D rendering, file compression, web browsing, floating-point calculations, and so on. After measuring CPU performance levels at each task, the numbers are weighted and combined into a single score.

The scores are “synthetic” because the tests used to calculate them are simulations. Rather than testing the processor’s performance in a specific application (say, a 3D creativity suite or game), they simulate the workload an application may place on a CPU under different circumstances.

In other words, synthetic benchmarks are not an exact predictor of performance. Instead they are widely used to compare the relative performance of CPUs. For example:

• PassMark runs heavy mathematical calculations that stress the CPU’s performance at compression, encryption, and physics-related tasks.
• 3DMark measures a system’s ability to handle 3D graphics for gaming.
• PCMark 10 scores the system on how well it can deal with business workflows and day-to-day productivity tasks.
• Intel® Extreme Tuning Utility (Intel® XTU) stress tests CPUs in addition to overclocking them and monitoring their function.

Each of these applications feature a database of scores ranking many CPUs, making it easy to compare the relative power of processors at a glance.

Real-world benchmarks

Use real-world benchmarks when you have specific plans in mind for your PC and need an accurate indication of performance for particular applications.

These tests are performed by giving real programs heavy workloads and then measuring the time it takes to complete. As a result, they provide a reliable preview of system performance when using the same settings.

Some commonly used applications for real-world benchmarking include:

• 7-Zip to measure a CPU’s data compression and decompression speeds.
• Blender to measure a CPU’s 3D rendering speeds.
• Handbrake to measure a CPU’s video encoding speeds.

In-game benchmark tools are another type of real-world test. These are non-interactive scenes that are available in some games. Use in-game benchmarks to check the CPU’s effect on FPS (frames per second) during regular gameplay and also while streaming.

These tests provide a repeatable test environment. As long as your system configuration remains the same, the benchmarks give an accurate reading of the in-game performance you’ll get.

What Your Benchmarking Stats Mean

Now that you know about the various kinds of benchmarking tests, let’s look at how to read the scores.

In synthetic tests, the scoring system will vary by program. Scores are often measured in “marks” (or another program-specific term). A higher-performance CPU is one that scores higher, though it’s important to remember different CPUs are designed for different purposes; not all are gaming-focused. Use a test that reflects the way you plan on using your CPU.

Real-world tests, on the other hand, use a range of different measurements.

• Dropped frames. In streaming benchmark tests, dropped frames counts the number of frames lost while encoding video. This can cause jumpy playback for viewers. A lower % of dropped frames is preferable.
• FPS (for video). In video encoding tests, FPS counts the number of frames your CPU encodes per second. Higher is better.
• FPS (for gaming). In in-game benchmark tests, FPS counts the number of frames rendered every second. A higher FPS usually means refined gameplay. (But frame time should also be taken into consideration).
• Frame time (1% low and 0.1% low). In in-game benchmarks tests, frame time (or frame pacing) counts the number of milliseconds between frames. Ideally, this measurement should be consistent. Otherwise, the frame rate’s pacing will be uneven, creating a stutter effect. When measured in milliseconds, lower is better. When converted to FPS to display alongside average FPS, higher is better.
• GB/s (gigabytes per second). In encryption tests, GB/s measures data throughput. Higher is better.
• MIPS (million instructions per second). In data compression tests, MIPS measures the number of low-level instructions executed by the CPU every second. Higher is better, but take the score with a grain of salt when comparing CPUs from different generations, as the means of executing instructions vary.
• Render time. In rendering benchmark tests, render time measures the speed at which your CPU renders the geometry, lighting, and textures in a 3D scene. Lower times are better.

Because some processors excel at certain types of CPU benchmark tests, it’s best to check several benchmarks instead of relying on a single number.

Synthetic and real-world benchmarks can complement one another. Check synthetic benchmarks for a big-picture view of the strengths of a specific CPU. Then use real-world benchmarks to get a better sense of how the CPU will perform in day to day use. Use both to fully understand the capabilities of a CPU.

When shopping for a CPU for gaming, for instance, use benchmark scores to evaluate a CPU’s general level of performance. Once you’ve found a few likely candidates, look up their FPS and frame times for some recent titles. (Or if you’re buying a CPU in anticipation of a game that hasn’t come out yet, you can look at some benchmarks for games built using the same engine and extrapolate.)

Single- Vs. Multi-Core Scores

Benchmarks are often split into single-core and multi-core scores.

Single-core scores are more relevant for games and applications that are lightly threaded, meaning they rely on a single core to process many — but not all — instructions.

Multi-core scores are more relevant for games and applications that are heavily threaded, meaning they distribute their instructions between multiple cores.

How can you tell if a game is lightly or heavily threaded?

• Open Task Manager. With the game running, open Task Manager (CTRL+SHIFT+ESC) in Windows 10 and click on the Performance tab. You’ll see an active graph for CPU utilization.
• Configure the CPU graph. Right-click the graph, then click “Change graph to > Logical processors.” You’ll see the load on each CPU core displayed separately.
• Compare core activity. If the game is lightly threaded, the majority of the activity will be isolated to one core.

Although quite a few popular games including Fortnite² are lightly threaded, more games are utilizing more cores. Heavily threaded games such as Assassin’s Creed Origins³ and Battlefield V⁴ can scale up to stress multiple cores. They will potentially have a higher FPS on a CPU with multiple cores.

System Configurations

While CPU benchmarks are important, every component plays a role in system performance.

• CPU. Games with complex AI, physics, and graphical post-processing tend to be more CPU-intensive and may benefit more from a CPU with a higher core/thread count and higher clock speed.
• GPU. Check GPU benchmarks in addition to CPU benchmarks when evaluating your system’s gaming performance, as some games are more reliant on the GPU. Discrete graphics cards can handle most of the work of 3D rendering, for instance.
• Memory and storage. These components can impact system responsiveness and loading times.
• Software. Regardless of your system configuration, performance will fluctuate from game to game. It just happens. This has to do with how the game was programmed. The graphical settings and resolutions you are playing on will also affect performance.

See our article on how to prevent bottlenecks to learn more about balancing your system.

Improving Your Benchmark Scores

Whether it’s high-end gaming, live-streaming your gaming sessions, or being creative and productive on your PC, the Intel® Core™ brand offers CPUs with performance that scales to whatever you’re doing.

For more insight into how to choose the right gaming CPU, read more here.