From the lab to the desk
Through the 1970s the University of Utah was where a surprising amount of the field's founding work happened. Ivan Sutherland had been hired into the program David Evans built. The two of them taught an advanced graphics class. Students from that room later founded Pixar, Silicon Graphics, and Adobe. Tom Stockham's image-processing group worked next door to the graphics lab. The split between "a picture as data" and "a picture as a scene" was already a corridor, not a gulf.
A hand, a face, a teapot
Edwin Catmull had come from Boeing and was taking a physics degree. He had grown up on Disney, loved animation, and had found out that he could not draw well enough to do it by hand. He treated the computer as the next step in animation rather than as a separate trade. The first computer animation he saw was his own: a hand opening and closing. In 1974 he pioneered texture mapping, painting a pattern onto a three-dimensional model. That is now one of the basic techniques. His stated goal was a feature film made of computer graphics. He got it two decades later, after a founding role at Pixar.
In the same class Fred Parke animated his wife's face. The hand and the face both appear in the 1976 feature Futureworld. They had originally been made for a 1972 experimental short, A Computer Animated Hand. A film audience met the pictures four years after a laboratory had finished them, which was about the normal delay.
Martin Newell's Utah teapot, and the still renderings of it, became a shorthand for CGI in the 1970s. A teapot is a good test. It has curves, a thin handle, a spout, and a body that has to shade smoothly or it looks like crockery from a bad dream. People are still rendering it, mostly as a joke that is also a measurement.
John Warnock was in that same circle. He later founded Adobe and, with PostScript, changed publishing: a page description a machine could carry from a screen to a press without the layout falling apart. Adobe went on to Photoshop, which became the standard tool for editing photographs, and to After Effects, widely used for motion work in film. James Clark was there too. He founded Silicon Graphics, whose machines dominated high-end rendering into the early 1990s.
Hidden surfaces, and shade that is not flat
A major practical advance at Utah was hidden-surface determination. To draw a 3D object on a screen the computer has to decide which surfaces are behind the object from where the viewer sits, and leave them out. Otherwise every back edge shows through, and the picture reads as a wire model even when you meant a solid. The 3D Core Graphics System, usually just called Core, was the first graphical standard in this line. Twenty-five people from SIGGRAPH drafted it as a conceptual framework. The specification was published in 1977 and became a base for later standards.
Shade arrived in the same decade, and it is what stopped the pictures looking like painted polygons. Henri Gouraud, Jim Blinn, and Bui Tuong Phong built the models that carry their names. Gouraud shading and Blinn–Phong shading let a surface suggest depth instead of a flat fill. In 1978 Blinn introduced bump mapping, a way to fake an uneven surface without adding geometry. A great deal of later mapping is a descendant of that cheat, and it is a useful cheat: the eye accepts a normal that wanders, even when the polygon underneath is flat.
Home computers were beginning to grow pictures of their own by the end of the decade. The earliest known 3D graphics program for a home machine is 3D Art Graphics, a set of effects written by Kazumasa Mitazawa and released in June 1978 for the Apple II. It was not a modelling package in the later sense. It was proof that the subject had left the lab, at least as a demo you could buy.
The arcade, then the chip
The modern arcade was born in the 1970s on real-time 2D sprites. Pong, in 1972, was one of the first cabinet games to become a hit. Speed Race in 1974 moved sprites along a road that scrolled vertically. Gun Fight in 1975 had human-looking animated characters. Space Invaders in 1978 put a large number of animated figures on one screen. Both of those last two used a barrel shifter built from discrete chips so that an Intel 8080 could move framebuffer graphics at a playable speed. The picture was already too heavy for the processor alone. That fact is the whole of the next decade, in miniature.
The 1980s are when the field stopped being an academic subject that happened to have a few products, and became a business with a large number of developers. Home computers spread. People who would never have seen a TX-2 were writing pictures.
Metal-oxide-semiconductor very-large-scale integration made 16-bit processors available, and it made the first graphics processor chips. NEC's μPD7220 was the first of those chips, a fully integrated NMOS part. It supported resolutions up to 1024 by 1024, and it was used on a number of cards. Intel's 82720 was a licensed clone, and Intel's first graphics processor. Memory got cheaper at the same time. Texas Instruments introduced video RAM in the middle of the decade, framebuffer memory that a display could read without fighting the processor for every byte. In 1984 Hitachi released the HD63484, the first CMOS graphics processor, capable of high-resolution colour and, in monochrome, of what would later be called 4K. In 1986 Texas Instruments introduced the TMS34010, the first fully programmable MOS graphics processor. A display controller that can run a program is a different object from a display controller that can only fill a rectangle you describe from outside.
There is a later argument about the word GPU. NVIDIA, in 1999, used it for a chip that did transform, lighting, and rendering together, and called the GeForce 256 the first. The μPD7220 is earlier, and it is a graphics processor in the plain sense. The two "firsts" are about different jobs. NVIDIA's claim belongs with the consumer 3D card.
Workstations, a Japanese ray tracer, a stained-glass knight
Terminals in this decade became intelligent. More of the graphics work moved onto the workstation and off the mainframe. A clear example is the Orca 1000, 2000, and 3000, built by Orcatech of Ottawa, a spin-off of Bell-Northern Research led by David Pearson. The Orca 3000 used a Motorola 68000 and AMD bit-slice processors, ran Unix, and was aimed at the expensive end of computer-aided engineering. Artists, meanwhile, were deciding that a personal computer could be a design tool, particularly the Amiga and the Macintosh. The Macintosh stayed in graphic design studios long after the novelty wore off. It was accurate, and it was faster than the previous methods, which is the only test that matters in a studio with a deadline.
Machines from the 1980s onward commonly used a graphical user interface: symbols and pictures rather than a command line as the way data is presented. Graphics had become one of the five key elements of multimedia, not a special effect bolted on the side.
For realistic rendering, Osaka University built the LINKS-1 in 1982, a machine with up to 257 Zilog Z8001 processors. The core of drawing a 3D image is working out the brightness of each pixel from the viewpoint, the lights, and the objects. LINKS-1 did that by letting each pixel be ray-traced independently, in parallel. With software written for that arrangement it could render highly realistic images quickly. As of 1984 it was the most powerful computer in the world, which is a useful reminder of how expensive a convincing picture was. In 1986 David Immel and James Kajiya set out the rendering equation, a step toward global illumination, which is what you need if the goal is a picture that behaves like light rather than like a painting of light.
Film work in the same years was pulled along by Star Wars and by other science-fiction productions. Lucasfilm and Industrial Light & Magic became the house other studios called. Chroma key, the various forms of "bluescreen," advanced across the later films of the first trilogy. Two pieces of 1985 outlasted the decade as markers. Dire Straits' video for "Money for Nothing" was near enough a fully computer-generated film to make CGI familiar to a music audience. A scene in Young Sherlock Holmes put a stained-glass knight on the screen, the first fully CGI character in a feature film. In 1988 the first shaders, small programs whose job is shading and nothing else, were written at Pixar, by then a separate company from ILM. The public did not see what those programs could do until the next decade. Late in the 1980s, Silicon Graphics machines were used at Pixar for some of the first fully computer-generated short films. For that decade an SGI workstation was what "serious graphics" looked like in a room.
The same years are called the golden age of video games, for a reason that has little to do with research papers. Atari, Nintendo, Sega, and the home computers, the MS-DOS PC, the Apple II, the Mac, the Amiga, showed computer pictures to a young audience that had not asked for them and then refused to give them back. If you could program, several of those machines let you write your own game. In the arcades, real-time 3D hardware arrived in 1988 with the Namco System 21 and the Taito Air System, the first dedicated boards for that job. On the professional side, Evans and Sutherland and Silicon Graphics built 3D raster hardware that leads, in a fairly straight line, to the later single-chip GPU: a powerful chip beside the CPU, doing the graphics work in parallel.
The decade also pushed graphics into markets that are easy to forget if you only watch films. The Evans and Sutherland Digistar projected stars for planetariums. Vehicle design, vehicle simulation, and chemistry all took the same methods and pointed them at a decision rather than at an audience. Donald Greenberg, at Cornell, spent these years teaching and writing. He founded the NSF Center for Computer Graphics and Scientific Visualization, and a remarkable number of his students, among them Robert L. Cook and Marc Levoy, went on to the work that came after. A field that can name its teachers that clearly is still young.