Benchmarking is a good way to determine how computer hardware performs compared to other brands or models. This can be helpful, especially if you are trying to figure out what you want to buy for your PC. Some examples are: processor benchmarking, where you compare its speed to other processors; graphics card benchmarking, which shows its speed compared to others in frames-per-second terms; power supply benchmarking, which tries to compare each power supply’s efficiency, and so on.
There are 2 main types of benchmarking: synthetic and real-world benchmarking. There’s kind of a fine line between the two, but looking carefully at them, you could determine what type of benchmark you are doing or looking at. I’ll try to explain as simple as I can to make the concepts more understandable.
Let’s begin with Real-world benchmarking as this could lay the foundation to better understanding the difference between the two. Real-world benchmarking aims to provide comparison of results of different hardware components in real-world scenarios or conditions. A perfect example of this is comparing frames-per-second values between graphics cards on a particular game. Another example is testing how fast a drive can read or write files, which is measured in seconds or minutes depending on the file size or testing time length.
Real-world benchmarking tests the totality of the performance of the hardware components being evaluated, and then compares them with one another. This type of benchmarking could allow you to confidently conclude that this particular hardware is a better performer than the other. Another way to look at it is how it will perform and what you’ll actually experience while you’re doing the same thing. If the benchmark says that it runs 75 frames per second on a particular game, if you have the same hardware in your computer, you’d probably be experiencing more or less 75 frames per second as well.
Synthetic benchmarks are tests that are specifically designed to capture a specific result for a specific aspect of the hardware, which you don’t normally perform or experience on a computer. Let’s say you want to test the rendering power of a video card, you’ll want to develop a test where you display textures and see how many of these can a graphics card process in a minute. Here’s another example. Testing the memory speed of a graphics card is a synthetic benchmark. This is because even if the memory speed of a graphics card is faster, it doesn’t mean it will perform better than another card with lower memory speed. The second graphics card could have higher bandwidth overall and/or faster graphics processor speed.
Synthetic benchmarks test the theoretical performance of the hardware. My understanding is that synthetic benchmark results can complement the results of a real-world benchmark. Synthetic benchmarks could tell you what the specific strength or weakness of a particular hardware is compared to the others in the same test. Another sign that it is a synthetic benchmark is that it is a program that is specifically designed for benchmarking or testing. You can’t actually play a game or process graphics and use it. It’s a set of repeatable tests that you could perform under different hardware setups for comparison.
And I guess the question now is which is more reliable or where you would most likely base your decision in buying a PC component from? Well, if you’re after the overall performance, looking at real-world benchmarks is much reliable since this is how the component will actually behave while you’re using your computer under real-world conditions. You would probably depend on synthetic benchmarks only if you want a specific aspect of a component aside from overall performance such as heat, power requirements and the likes.
Relying on one benchmark doesn’t necessarily reflect the actual performance comparison of the hardware in a particular list. It is when you compare the results from different benchmarks that you’ll get a more reliable performance comparison between the hardware you are trying to evaluate and other hardware. Real-world and synthetic benchmarks test hardware differently, but having both data can help in understanding the performance differences of the hardware components in the benchmark. Ultimately, you can then conclude on which of them is better or the best performer and whether the hardware you are looking to buy is a good purchase or not.