Why is processing speed measured in cycles?
Typically, when you think of something cycling, you think of it spinning until it’s made it all the way back around. Processing speed, on the other hand, is measured in cycles while nothing spins. Since there aren’t any moving parts in the silicon, why use that terminology?
Well, it’s because processors go through a sequence of events to complete a task. Some people explain it by saying that a processor can execute one instruction per cycle, but the sequence is more formally known as a fetch execute cycle.
The fetch execute cycle
The fetch decode execute cycle determines speed of any computer component measured in hertz.
In layman’s term, “fetch” happens when a component receives an instruction, “decoding” is when it figures out what the instruction means and “execute” is when the input is completed. At a basic level, this sequence of events would represent one cycle or 1 Hz.
Unfortunately, not all clock speeds are made equally. Some instructions require more than one clock cycle to execute and others can be completed multiple times within a cycle.
Processor architecture performance evaluation
A more accurate way to measure performance is by calculating cycles per instruction. I mentioned earlier that different types of instructions can take more or less than one cycle to complete. CPI is the average number of processing cycles per instruction.
You can see how the math is broken down in this video:
While this gives you a more detailed perspective, it is not information readily available on consumer processors. Also, comparing CPI to CPI is only relevant if both processors are using the same architecture.
There are many more metrics that can be used to compare performance like this, like MIPS, MFLOPS and SPECint but at the end of the day, aggregate ratings may not reflect what your experience will be like.
How should you actually gauge CPU performance
The most important metric for CPU performance is how efficiently it performs in the tasks that you use it for. When you look at gaming benchmarks with a Ryzen 3900x faced off against an i9 9900k, they end up trading off wins per game and depending on which settings the game is running at.
It’s an uncomfortable reality for those who would drop tons of money to have the absolute best performance across all aspects. There is simply just a single component that is best across all use cases.
Once you understand this, you’ll be able to better navigate computer component marketing. Don’t buy something because it has a higher clock speed, don’t buy it because it has more cores. Buy it because you’ve seen someone test it in the scenario that you will most often use it for and it provided good results.
