When buying and picking your CPU, you want to get the best you can get for your money, as well as the CPU that best fits your needs. When sifting through the available CPUs to satisfy those goals, you are bombarded with numerous technical terms and specifications.
In this article, I aim to clear up what those terms mean, and (roughly speaking) how their specs translate into performance.
Important CPU Specs
Core Count and Thread Count
First up is core and thread count; this is one of the most widely marketed specifications when it comes to processors. Generally speaking, a processor does calculations, and every core does one calculation at a time. The more cores a CPU has, the more calculations it can do at the same time. However, most cores can do two at the same time nowadays; this is called SMT (AMD) or Hyper-Threading (Intel).
This means that processors with hyperthreading have the same calculation power, but are more efficient. Without SMT/Hyper-Threading, cores continuously switch between calculations to do them all at the same time. The time it takes to switch is lost performance. With hyperthreading, cores don’t have to switch as often; they spend more time calculating, less time switching, and thus perform better.
These types of performance considerations are affected by additional cores as well as additional threads. So, the more cores and threads your CPU has, the better it might do at complex tasks. But this simple rule of thumb is made more complicated by two things: (1) software has to be purposefully designed to take advantage of high core/thread counts, and (2) in many program situations there are tasks that simply have to be done in a certain order, so clock speed will often be a bigger determinant of performance than core count.
This brings us to the other ‘most famous’ CPU specification: clock speed. This is the speed at which your processor can receive and interpret new instructions/calculations. Clock speed is often referred to as ‘single-core performance’; this is because each core is individually clocked at that speed (unless, of course, some cores are clocked higher because the CPU has a ‘boost’ feature which allows some or all cores to run at higher speeds).
Practically speaking, the higher the speed, the faster your cores solve calculations. Many chips can be overclocked (given a sufficiently capable CPU Cooler), but most users stick with the stock speed. Some programs will benefit greatly from higher clock speed, such as emulators and video editors. Other workloads, like office multitasking and conventional gaming, might not benefits as much (i.e. as noticeably) from higher clock speeds.
You may notice at some point that certain CPUs with the same number of cores, same number of threads, and same clock speed still perform differently. This is usually because of IPC (Instructions Per Cycle). Clock speed tells you how many processing cycles happen in a second, but IPC tells you how many instructions are received and interpreted in each cycle. Newer chips have a higher IPC and, as such, can work faster than older chips—even when some of the older chips may be clocked at higher speeds!
Sometimes (rarely) instruction sets are also mentioned; these are the different ways a processor can receive instructions for calculations. The more optimized these sets are, the faster your chip is. However, the performance bump is so small, and the instruction sets have been optimized so far, that you can safely ignore this when making practical decisions between different processors for your build. It would be very unusual for this to be an especially important specification for any given CPU that you’re considering.
The CPU cache is where the CPU stores its calculations and data for those calculations. This memory is located very close to the CPU, so that it can be accessed at the extreme speeds of processing cores. Otherwise, the CPU would have to use RAM, which (although much faster than most hard drive storage) is far slower than CPU cache. The cache is divided into ‘levels’. L1 is the closest to the CPU and the fastest. L2 and L3 are further away, but often bigger; this means lower speeds, higher capacity.
More cache generally does mean more performance, since a CPU can receive and temporarily store data faster—but the performance difference for this attribute between processors in any given generation of CPUs still tends to be minor when compared to the differences introduced by core/thread count and clock speed.
The Thermal Design Power or Thermal Design Point of a CPU is the maximum amount of power that the chip is rated to generate as heat while in operation, measured in Watts. So, if this figure is very high, that means the CPU uses a lot of energy and generates a lot of heat, and that it will require not only a fair amount of power from the PSU—but also a fairly strong cooler to keep it at a low temperature. Generally speaking, TDP is a specification that will matter more to users that care about power efficiency and/or system noise moreso than those that care about performance above all else.
Lithography or Process
With the release of new chip generations, lithography is often mentioned: “AMD’s 7 nm chips” or “Intel’s new 10 nm chips”. This measurement in nanometers stands for how big the transistors are. The smaller these are, the more you can fit on a chip die; moreover, generally speaking, smaller transistors consume less power and thus run cooler. However, their size tells you nothing about the performance of the chip.
It does, of course, tell you that they managed to fit more transistors on the die, but it would be a better idea to check out other specifications if you want to know about the performance of a chip. The reason I still included lithography or process is that it is so frequently mentioned and reported. Sometimes you will see it (inaccurately) called ‘architecture’, which is a general term for the layout, design, and low-level functionality of a piece of hardware.
Hopefully, this has cleared some things up for you, and helped you on your journey of picking your next CPU! But if anything is still unclear, please feel free to ask about it. This is just a quick overview of key CPU specifications, after all.
And if you’re new to the whole concept of building a PC, you may want to take a peek at our quick 4-minute overview on the broad strokes of the assembly process: