Both CPU (Central Processing Unit), also known as “processor” and GPU (Graphics Processing Unit), most commonly known as “graphics card” are essential parts of a modern computer. They work in unison, and a good combination ensures smooth performance. But since these components aren’t free, and most people cannot afford to purchase them at once, it’s important to see which one should be a priority based on your needs. Let’s get into a battle of CPU vs GPU, although it isn’t truly a battle – they need each other to function.
What are CPUs and GPUs made of?
When you get to the bottom of it, the CPU and GPU are very similar. They’re built out of silicon and are of similar size, but designed for different purposes. Neither can be fully replaced by the other – they are built to complement each other. The CPU has a larger number of control units and a bigger cache, but a smaller number of ALU (arithmetic logic unit). GPU has a massive number of ALU, but a very small number of control units and cache.
In some cases, CPU comes equipped with an integrated GPU. Although much weaker than the full-size alternative, it has its use. The primary use is monetary and very beneficial for companies that need large quantities of computers for their employees. They only need to buy a CPU, and the internal GPU is strong enough to handle video output to their monitors, and perform light tasks. But, the technology improvement in the early 2020s has also benefitted those on a budget. Some internal GPUs are now used for entry-level gaming with greatly reduced graphics settings or less-demanding games.
The term “size” can mean a lot of things. We’ll mention one of the aspects under “Cooling” at the bottom. But what is most associated with is the word size in CPUs – 32-bit used to be common but is hard to find nowadays, as 64-bit word size CPUs have overtaken the world. Also, CPUs have a cache size, separated into three levels (L1, L2, and L3). With GPUs, the size can refer to memory bandwidth – 128-bit, 256-bit, and even 384-bit. Neither of these features of CPUs and GPUs is directly comparable. What is, however, is the clock speed. Consumer GPUs have a much lower clock speed, typically between 500 MHz and 2000 MHz. Consumer CPUs, on the other hand, typically operate at a clock speed between 2000 MHz and 5200 MHz.
Functions of CPU and GPU
GPUs are designed to handle parallel processing/computing and consist of hundreds of small cores that have thousands of threads. They all work simultaneously, provide high throughput, and can handle thousands of operations at once. CPUs, on the other hand, are designed to have low latency and a small number of cores. The number isn’t small – between 2 and 48, in most cases. Consider the fact that the majority of CPUs also have Hyper-Threading, which doubles the number of cores as threads. For example, AMD’s Ryzen Threadripper 3970X has 32 cores and 64 threads.
To simplify the matter, think of the CPU as the brain. It can handle one or a smaller number of tasks very efficiently. GPUs can handle a large number of operations within one clock cycle, but are “dumber” and suited for repetitive computing tasks. Also, CPUs can integrate with, and organize virtual memory, and come equipped with certain instruction sets. Let’s clarify – CPUs can handle the input and output of your computer’s components, control data storage, relay commands to other components, write data, and control the timing of information flow.
Possibly the easiest way to spot a difference between CPU and GPU is when you use them for the task they weren’t designed for.
Speed of processing tasks
GPUs have more memory bandwidth and more powerful processing power, which is why they’re used for machine learning, AI, statistics, generating 3D graphics, and analyzing and calculating large amounts of data. The most famous use for GPUs is playing video games and video, image, and animation rendering. But, they are used for a wide array of scientific, engineering, enterprise, and consumer applications. Remember when people built GPU “farms” to mine cryptocurrency such as Bitcoin?
From what you’ve just read, it’s easy to conclude that GPUs have a faster render time. After all, the time to tackle certain tasks can be up to 100 times faster in theory. In reality, the time saved won’t be as high, but still significant. This is why, with the popularity of Virtual Reality, the system requirements in terms of GPU are quite high – they need to render all graphical elements, and fast!
But we also need to mention one possibility. Although slower, CPUs tend to produce images of higher quality, and with less noise. So if time and budget weren’t a problem, it would be better to “interconnect” several CPUs for a better outcome. This is a technique used by film studios in Hollywood when the demand for quality is very high.
So, how does GPU vs CPU fare in terms of cooling? Both CPU and GPU have a die – a block of semiconducting material that contains the integrated circuits. CPUs have an IHS (Integrated Heat Spreader), a metal “lid”, over the die. Either liquid metal, thermal paste or a carbon pad is placed between the die and the IHS. That layer transfers the heat between them. Another layer of paste, pad, or metal is applied between the IHS and the external cooler on CPUs, which dissipates the heat through air or water.
GPUs are much bigger primarily because the die is placed on a larger PCB, which contains other components that assist the GPU. But it is also because GPUs require better cooling. Once again, a carbon pad, liquid metal, or thermal paste is applied on the die, and a heatsink, typically made of copper or aluminum, is placed on top of it. Those metal are known for thermal conductivity and carry the heat away from the die, which is then dissipated by one, two, or even three fans. Or, the GPU can be modified with a water cooler system, which uses a pump to cycle water and carries the heat away.
If GPU and CPU aren’t balanced, this causes a bottleneck – the term used when the weaker components start dragging the entire computer’s performance down. In that case, consider replacing the older component. This isn’t failsafe, because a newer entry-level component is weaker than an older high-end one. You can also download overlay software such as MSI Afterburner. It will let you know, in real-time, the percentage of CPU and GPU that is used. If one of them is stuck on 100% and struggling, and the other is at 30%, you spotted your problem and know what needs to be upgraded. Another solution is to search for benchmarks of a system using your current components or those you’re planning to buy.
You maybe also interested to know which one of the SSD or HDD Is better.