What is a GPU and how does it work?

The GPU is created in a very different way to the CPU.

Graphics. They immerse us in worlds, tell us rich stories and make us gawk at insanely realistic simulations. That’s all thanks to powerful GPUs—Graphics Processing Units. GPUs are one of the most amazing parts of modern computers. While CPUs can be thought of as the brain of the computer, GPUs are a kind of visual artist. Taking the information from the CPU and translating it into something visual. But how does it do that? And why exactly are GPUs so expensive? Let’s have a closer look.

Inside the GPU

ADVERTISEMENT

The GPU is created in a very different way to the CPU. While the CPU handles tasks sequentially, the GPU is specialised for parallel processing. Meaning that it can break down complex tasks—particularly graphical calculations—into smaller chunks and process them simultaneously across its many cores. This is similar to how CPUs use multiple cores to handle multiple tasks; but supercharged. And those specialised tasks being the calculations required to manipulate textures, lighting and rendering. Particularly 3D.

GPUs also have something called VRAM. This is dedicated video memory, which acts as temporary storage for the vast amounts of input data the GPU needs to render scenes in 3D. This VRAM is crucial to the running of a GPU, providing minimal latency so that the GPU can respond in real time.

From Data to Display

Let’s take a closer look at 3D rendering on the GPU. Following it on its journey from creation to display. Imagine we are in a modelling software like Blender. This is roughly how the GPU will render what you see—albeit in a VERY simplified way.

1. Modelling. We start with the model itself. Defining the physical properties such as vertices, polygons, textures and the rest. These parameters provide valuable inputs which can then be drawn by the GPU.

2. Storing the Model. Once the model is in; its data is sent from the CPU to the GPU’s VRAM, storing those parameters for rendering reference. If the model changes, the CPU can easily update these parameters.

3. Vertex processing. Now the magic starts. The GPU’s vertex processors take the information from the VRAM and use it as instructions to draw the model. Allowing it to calculate everything from position to scale.

4. Rasterization. While everything has been in 3D up until now, our screens are all 2D. Meaning the image must be rasterized—turned from 3D coordinates into the pixels which show on the screen.

5. Fragment processing. The GPU then takes these rasterized elements and applies textures, lighting effects, shadows and other enhancements on a pixel-by-pixel basis. Turning the image into a living, breathing scene.

6. Frame buffering. Finally, the processed pixel data is stored in the frame buffer within the VRAM. Ready to be rendered on screen.

7. When ready, the frame buffer then sends all the information to the display adapter. Translating the precise pixel information to render on the target display. Letting you see the output.

But the GPU’s task isn’t over. Say you’re playing World of Warcraft. This process must be done anywhere from 30 to 240 times every second, to provide smooth visual outputs for animation and gaming. Fortunately for us, we don’t have to think about it. So we can get back to getting the best loot and gear in WoW, and buy wow cataclysm gold to make the process even easier.

Beyond Gaming

GPUs are clearly very important for rendering in video games. But their use cases extend well beyond just that. Let’s take a look at a couple of examples:

• Science: Research often requires simulations. This can range from protein folding to climate modelling. The power of parallel processing offered by GPUs can significantly reduce computation time for scientific simulations.
• Machine Learning: Training AI models relies on tonnes of computation. Seriously insane amounts of it. Again, as GPUs offer parallel computation, they are great for training AI models faster than CPUs.
• Video Work: Rendering isn’t just something needed for games. Working with large video files also requires rendering and encoding large amounts of data—allowing GPUs to help speed it up.
• 3D Design: Of course, rending virtual worlds is not something limited to video games. Architecture, art and simulations all require careful 3D design and rendering. Thus powerful GPUs are the backbone of any 3D rendering.
• Crypto: Crypto mining requires solving a lot of arbitrary math problems at lightning pace. While CPUs are good for that, again it’s the power of parallel processing which makes GPUs such a useful tool for it.

While each of these workflows will likely use a slightly different method of operating the GPU. The same key parts of using the cores for simultaneous processing and the VRAM for storing parameters or outputs will continue across all of these use cases.

The Price Tag

So GPUs are useful. And, now, almost essential for a load of different activities. But getting your hands on a powerful one can be tough. Mainly due to their price tag.

After understanding how the GPU executes its tasks, and the types of tasks those are, it’s quite easy to see why GPUs would be expensive. They require lots of powerful parts which work simultaneously to complete complex calculations at lightning speed. But there are three key reasons that GPUs are super expensive compared to some other parts.

1. Manufacturing Complexity. These components are seriously complex. And thus require some insanely complex manufacturing processes to place the billions of transistors onto their silicon. It’s expensive, specialist and virtually impossible outside the world’s top GPU manufacturers.

2. High Demand, Limited Supply. You may remember GPU shortages during the COVID-19 pandemic. These were due in part to the explosion in crypto mining causing GPUs to be bought out. And as the manufacturing is so complex, it’s hard to scale up production in short periods.

3. Research and Development: if you didn’t get it yet; GPUs are insanely complex. And producing new GPUs requires significant investment by for-profit companies. And that research is something that consumers of the GPUs ultimately have to fund through their purchases.

The Powerhouse of Modern Computing

Being the heavy lifter of rendering, simulation, AI training, machine learning and more GPUs are really the modern powerhouse of the computer. Because of this, we’re likely to see more and more development into powerful GPUs, increasing their speed and power.

Overall, GPUs are complex. But they are the powerful components behind rendering compelling virtual worlds.

"Exciting news! Mid-day is now on WhatsApp Channels Subscribe today by clicking the link and stay updated with the latest news!" Click here!