Research before products
The first interactive computer graphics were built in laboratories in the early 1960s. Ivan Sutherland’s Sketchpad, developed at MIT in 1963, let a person draw on a display with a light pen. In the same decade, Sutherland and others began working on three-dimensional displays, including a head-mounted system often called the Sword of Damocles. Early machines had to calculate every point and line on an ordinary processor, which was slow, so researchers soon designed special hardware for the repetitive geometry involved.
Evans & Sutherland and flight simulators
In 1968, David Evans and Ivan Sutherland founded Evans & Sutherland, a company that built graphics systems at a time when the market was small and expensive. Their customers included flight-simulator makers and the military, who wanted moving scenes drawn in real time. Hardware of this kind could handle the transformation of 3D shapes, clipping and perspective much faster than a general computer. These systems cost enormous sums, and few people outside industry or research ever saw one.
The workstation era
In the early 1980s, engineering workstations from several companies placed a powerful computer and a graphics display on a single desk. Silicon Graphics, founded in 1981 by Jim Clark and colleagues, became a leading name. Its Geometry Engine, a custom chip developed at Stanford in the early 1980s, performed the core mathematics of 3D drawing in hardware. Silicon Graphics machines became widely used for computer-aided design, scientific visualization and film effects.
The company also promoted a programming interface, IRIS GL, which was later reshaped into the open standard OpenGL in 1992. Having a shared interface mattered: software could use whichever hardware was available, and the standard helped spread the techniques beyond one vendor. Microsoft’s Direct3D later became the other major interface on Windows.
Why the leap to consumers took time
Workstation graphics relied on custom chips and costly memory. Several things changed in the 1990s. Chip manufacturing improved, memory prices fell and more households owned a PC with enough power to benefit. Game developers, particularly those making 3D action games, wanted more than the main processor could provide. Dedicated cards from companies such as 3dfx, ATI and Nvidia answered that demand, bringing ideas first tried in expensive systems to a price ordinary buyers could pay. Many early 3D companies disappeared or were acquired, since competition was intense and product cycles short.
Shaders and programmable hardware
Early consumer cards used fixed functions: a given operation was wired into the chip. In the early 2000s, programmable shaders arrived. They let developers write small programs that ran for every vertex or pixel, giving more flexibility in lighting and surface effects. This programmability proved significant. A graphics processor carries out the same operation on many pixels at once, so it is built from a large number of simple units working in parallel.
Beyond graphics
Researchers noticed that this parallel design could be used for other calculations, such as physics simulations and linear algebra. Early experiments in the 2000s coaxed general computation from graphics hardware by disguising data as images. In 2007 Nvidia released CUDA, a programming model for general-purpose use of its processors, and other companies created alternatives. Scientists adopted these tools for simulation and data analysis.
When researchers working on neural networks found that GPUs sped up training dramatically, they became central to the rise of modern artificial intelligence. Cryptocurrency mining also drew demand for graphics cards in some years, which affected prices and supply, though that is a side note to the main history.
Specialized chips today
Today graphics processors are among the most complex chips made. Some designs now include dedicated units for ray tracing, which simulates light paths for realistic images, and for matrix calculations common in machine learning. Manufacturers also build separate accelerators for AI work that differ from graphics cards, showing the same pattern repeating: a specialized hardware idea starts in research, becomes a commercial product and then spreads to new tasks.
What to take away
The commercial graphics card was not a single invention. It grew from laboratory ideas, was refined in professional workstations and reached the mass market once manufacturing costs dropped and gaming created demand. Its second life as a parallel computing device was largely unplanned, a good example of how hardware built for one purpose can find another.