GPUs: Procesadores gráficos y de AI

GPUs(Graphics Processing Units) transformed computing by enabling advanced graphics and parallel processing, changing everything from early video games like Pong to modern artificial intelligence and cryptocurrency applications. Understanding how GPUs work shows us how graphics technology positively affected multiple industries and our daily lives.

What were the first computers and how did GPUs appear?

In the beginning, computers did not have displays, their results were printed on paper. The first displays came from the oscilloscope, an electronic instrument that displayed electromagnetic waves. Engineers adapted the oscilloscope and the first video game was born: Pong, marking the beginning of digital graphics before visual operating systems.

Later, simple games such as Tetris and Prince of Persia used minimal graphics capabilities on platforms such as UNIX or DOS. Thanks to this, computing took full advantage of the available chips driving a specific industry, GPUs.

What is the difference between CPUs and GPUs?

A CPU performs complex calculations, but in sequential cycles, suitable for everyday programs such as Excel or web browsing. A GPU, on the other hand, has many smaller cores that execute simpler calculations but simultaneously, ideal for generating graphics made up of millions of pixels at the same time.

For example:

• Serial processing (CPU): when an image loads slowly line by line.
• Parallel processing (GPU): when an image is updated quickly, ideal for video games and animation.

What are VRAM and CUDA and how did they drive parallel computing?

GPUs have a specialized memory called VRAM(Video RAM), designed to run parallel graphics tasks such as texturing in video games and complex simulations. This capability made scientific applications possible, creating supercomputers at lower prices as happened with PlayStation 3 and Linux.

NVIDIA's CUDA is an intermediate language that allows GPU programming directly for tasks beyond graphics, such as physics simulations, artificial intelligence and data processing in cryptomining.

How have GPUs influenced artificial intelligence and other industries?

Parallel processing on GPUs is also essential in:

• Artificial Intelligence: training models such as ChatGPT, where they simultaneously explore multiple possibilities to answer questions.
• Autonomous cars: simultaneously analyzing visual information to make quick and safe decisions.
• Cryptocurrencies: performing parallel cryptographic calculations that are basic to blockchain, leading to the creation of huge mining farms.
• Climate and scientific simulations: processing multiple variables simultaneously to model complex phenomena.

This led to huge growth in demand for GPUs.

What recent technologies have increased the need for GPUs?

Currently, technologies such as virtual and augmented reality are driving major graphics innovations. Each eye requires separate graphical representations, demanding intensive and efficient graphics processing. Valve developed SteamOS, optimizing PC resources for gaming and virtual reality.

Animation, special effects and audiovisual content companies also rely on advanced GPUs, especially equipped with specific units to process graphics algorithms such as ray tracing (physical simulation of light) and multimedia (video) codecs.

How does the hardware design of GPUs affect their performance?

GPUs have special hardware considerations. They are often connected directly to the motherboard via PCI-E ports. Due to the intense parallel processing, they require efficient dedicated cooling systems: fans, liquid cooling and even oil immersion, a technique used to effectively disperse heat for demanding tasks.

In mobile devices and Apple computers with System on a Chip, GPUs and CPUs are integrated on the same chips, increasing power efficiency and graphics performance.

We invite you to share your experience with the use of GPUs or any questions about their performance.