How a chip designed for video games became the engine of modern artificial intelligence
In short: Originally built to render 3D polygons for arcade machines, the graphics processing unit has evolved into a powerhouse of parallel computation. By offloading thousands of simultaneous tasks from the main processor, these chips now drive everything from complex neural network training to the most demanding scientific simulations in existence.
The modern graphics processing unit emerged in the 1990s, evolving from simple video display controllers that merely moved memory to sophisticated circuits capable of independent 3D calculations. While early hardware like the 1981 Geometry Engine laid the groundwork for vector processing, the term itself was popularized by Sony in 1994 for the PlayStation console. These early systems were designed to handle specific graphical duties, such as line drawing and texture mapping, without burdening the central processor.
The true revolution occurred when engineers realized that the parallel architecture required for rendering graphics—handling thousands of pixels at once—was perfectly suited for other data-intensive problems. Modern GPUs contain thousands of calculation units that excel at linear algebra. This capability transformed them into general-purpose engines, or GPGPUs, capable of running compute kernels for tasks like protein folding, cryptocurrency mining, and large-scale modeling. In some scientific applications, these chips perform calculations forty times faster than traditional CPUs.
Performance is typically gauged by the number of floating-point operations completed per second, often expressed in teraflops. However, this metric is an estimate, as actual speed depends on factors like memory bandwidth, clock frequency, and the specific architecture of the processor cores—whether they are Nvidia's streaming multiprocessors, AMD's compute units, or Intel's Xe cores. Today, these chips are essential to supercomputing, with the majority of the world's top-performing systems relying on GPU-integrated architectures to handle the massive, parallel workloads required for modern artificial intelligence.
The distinction between dedicated and integrated graphics remains a critical divide in hardware design. Dedicated cards feature their own high-speed memory, providing massive bandwidth for demanding tasks. Conversely, integrated solutions share system RAM with the CPU, which is more cost-effective but can create performance bottlenecks. Despite these differences, the shift toward unified memory architectures in modern consoles and processors continues to blur the lines, ensuring that the GPU remains the central component for both visual fidelity and computational power.
Opens the Intellect+ feed with this card first. Swipe up for the next one.
Sources
- Summary by Intellect · source: Graphics processing unit
To cite this card, link to https://intellectplus.co.uk/app/technology-graphics-processing-unit.
How we source, check and correct cards: Editorial standards.
General information, not professional advice. Spotted an error? Flag the card or see corrections and complaints.
Intellect Plus is a visual educational discovery platform for exploring fascinating facts and ideas. Brain Food, Not Brain Rot. What is Intellect Plus?
Related
More in Technology today · More in Technology · Latest · A–Z · All topics