Sign in

GPU

1999Big moment

Nvidia markets the GPU. Built for games, GPUs later power the AI revolution.

The story

On October 11, 1999, the American chip company Nvidia began selling the GeForce 256 and marketed it as the world's first graphics processing unit, or GPU. The chip took over the work of transforming 3D shapes and lighting them, calculations that until then ran on the computer's main processor, and it could process about 10 million polygons per second. Nvidia had been founded in 1993 by Jensen Huang, Chris Malachowsky and Curtis Priem, and it chose the term GPU to present the product as a new class of processor rather than a faster graphics card.

The basic idea came from the microprocessor, the general-purpose chip that had made personal computers possible. Drawing a 3D scene means doing the same small arithmetic on millions of pixels and vertices at once, a task that suits many simple processors working in parallel more than a few powerful ones, and the arrival of 3D games in the 1990s created a market large enough to pay for rapid development. Rivals such as 3dfx and ATI also competed, but Nvidia came to dominate.

The long-term surprise was that GPUs turned out to be good at many things besides graphics. Starting with programmable shaders in 2001, scientists began using them for general calculations, and in 2007 Nvidia released CUDA, a programming platform that made this practical. In 2012, a neural network called AlexNet, trained on two Nvidia gaming GPUs by Alex Krizhevsky, Ilya Sutskever and Geoffrey Hinton, won the ImageNet image recognition contest with an error rate of 15.3 percent, against 26.2 percent for the runner-up. This started the deep learning boom, and training the large models since then has depended on huge clusters of GPUs. Nvidia's market value passed 1 trillion dollars in 2023.

Why it mattered

  • The AlexNet result of 2012 showed that GPUs could train deep neural networks quickly, beginning the modern deep learning boom.
  • Large language models and image generators are trained and run mainly on GPU clusters.
  • GPUs became essential in scientific simulation, from weather forecasting to protein structure prediction.
  • Cryptocurrency miners drove surges in GPU demand and prices during several periods of the 2010s and 2020s.

Drafted by AI (Claude) from the sources above and checked against them by AI; no historian has reviewed this page. Spot a mistake? Suggest a correction. How pages are made

Community notes

Loading notes…

Learn more ranked by votes

BookTae Kim
The Nvidia Way
Search Open Library
BookStephen Witt
The Thinking Machine
Search Open Library
VideoSearch YouTube
history of the GPU Nvidia GeForce 256
Search YouTube