The Painter in the Computer


GPU Core

Very good at graphics, especially drawing triangles, she is the art of top IC technics. In the begining she does the rendering tasks instead of CPU. After generations of evolution, she can also share some parts of the calculation tasks of CPU.

On computers with GUI, after a program has done its work, the next step is usually to draw and display the results on the screen. In early computers, there were all the CPU's job. However, when the screen resolution is getting higher and higher, and users have the need to play games and videos on the computer, many computers have another IC board called "graphics card" installed into the computer, so that it can share the work of drawing the screen. In addition to video signal processing, most graphics cards have the function of drawing points, lines, and faces on the screen. So graphics cards are also called "3D accelerators" which are crucial equipment for running computer games.

From Graphics Adapter to GPU

Based on the points, lines, and faces data that the CPU sends to the graphics card, the graphics card's computational circuitry is able to project the points, lines, and faces in 3D space onto a flat 2D screen. It calculates the lighting and colors of the primitives and fills the pixels inside the shapes on the screen. These calculations are just fixed mathematical formulas, so the graphics card's computational circuitry doesn't need to change much. However, if the computational circuitry can be replaced with a processor that can substitute programs, it can change the formulas of the drawing process, and then can do more calculations, so the graphics card can be used to draw a variety of different styles of pictures, such as cartoon, ink, or more realistic game scenes. Now the graphics card has its own processor, called the Graphics Processing Unit (GPU).

After that, the development of graphics processing unit (GPU) is similar to CPU. A graphics processing unit (GPU) chip can contain many circuits called "GPU cores" which allow it to compute the different parts of an image simultaneously. Basically, GPU cores are simpler mathematical circuits than CPU cores, so a large number of GPU cores can be embedded in a GPU.

Today, most personal computers require both CPU and GPU features, so there are two types of configurations for CPU and GPU products. One is to embed a small number of GPU cores in a CPU chip, and the other is to keep the CPU cores in the CPU and install a graphics card with a large number of GPU cores. The former is usually used for affordable computers that do not pursue game quality, while the latter is used for entertainment computers or multimedia work computers that emphasize gaming or multimedia performance.

From Graphics to AI

In addition to gaming, the GPU's ability to perform a large number of simple mathematical operations at the same time makes it more suitable than the CPU for tasks that require a large number of parallel computations. One of the tasks is neural network-based artificial intelligence (AI) computation. Whether it's a researcher who wants to train a new AI model, or a general user who wants to use an existing AI model, GPU helps a lot. So the demand of GPUs in modern PCs is also becoming increasingly large.

See Also