A CPU core is an essential part of any computer.
CPU cores are a part of any CPU processor.
Its worth noting that cores arent entirely independent of each other.
Depending on the particular CPU design, cores can be more or less closely linked.
More often than not, cores will be connected via buses.
There is also a distinction between CPUs that have only identical cores and those that combine different ones.
Contents
CPU Design
Historical multicore CPU design has generally used a homogenous CPU topology.
That is, all cores are identical.
It also makes task scheduling easier as all cores can perform all tasks at the same speed and efficiency.
A more nuanced approach to CPU core design can be found with heterogenous CPU topology.
Historically, CPUs started with only one core and could only handle one task at a time.
Over time, as demand for hardware increased, that wasnt enough anymore.
Newer, more modern CPUs were developed and phased out than those with fewer cores.
Multithreading
Most modern processors use multi- or hyper-threading to increase the number of cores available.
This process splits one core into several virtual cores.
Specifically, each physical core works as two threads.
Therefore, CPUs with four cores can work with eight threads, meaning they function like an eight-core CPU.
Note:Some specialized CPUs can offer more than two threads per CPU core.
However, all such products are exclusive to the HPC (High-Performance Computing) and supercomputing markets.
Desktop CPU cores can either run one or two threads.
Multithreading isnt an absolute duplication of CPU power, though.
Hyperthreading does not double the performance of the CPU core.
Some apps and programs work with it better than others.
Other software, especially video editing and animation, runs farther with extra cores and threads.
They can be used by whichever thread needs them more.
Theoretically, it would be possible to build CPUs with hundreds or even thousands of cores.
That isnt yet a commercial reality, with AMDs Threadripper and EPYC CPUs having up to 64 cores.
For now, though, a more realistic focus is optimizing the performance-per-watt.
In other words to lower the power draw of CPUs.
This primarily benefits laptops and other battery-powered devices.
Managing power consumption is critical for further significant performance increases.
Moores law has generally doubled CPU performance roughly every two years for decades.
Modern CPU nodes are so small that theyre very close to physical limits on size reductions.
Increasing performance has therefore meant higher power draw and higher heat output.
Naturally, there are also always new types of CPUs being developed.
The two biggest brands here, Intel and AMD, each boast different types of CPU designs.
This goes so far that their respective CPUs are better suited to some uses than others.
Of course, newly designed CPUs offer new use cases and specialties on top of existing ones.
CPU architecture is a complex topic.
Like the GPU market, the CPU market shows signs of shifting towards specific hardware accelerators.
This can allow more performance and greater efficiency at particular tasks but increases complexity.
These will typically be served and surrounded by registers and caches.
The vast majority of modern CPUs offer multiple cores on one CPU die.
CPU cores may be identical or optimized for different stages on the performance/efficiency curve.
CPU cores are typically general-purpose, able to perform any processing the CPU might need.
A non-general-purpose processing unit on a CPU die may be called an accelerator or an X processing core.
X is replaced with a specific purpose, such as neural processing cores and neural accelerators for AI processing.
