Tom explains what GPUs are, why they are different than CPUs, and how they kinda just evolved into a product.
Featuring Tom Merritt.
Please SUBSCRIBE HERE.
A special thanks to all our supporters–without you, none of this would be possible. Become a supporter at the Know A Little More Patreon page for exclusive content and ad-free episodes.
Thanks to Kevin MacLeod of Incompetech.com for the theme music.
Thanks to Garrett Weinzierl for the logo!
Thanks to our mods, Kylde, Jack_Shid, KAPT_Kipper, and scottierowland on the subreddit
Send us email to [email protected]
Episode transcript:
Ask a gamer about hardware and they will spend a lot of your time talking about the GPU. Ask someone about GPUs and they’ll spend a lot of time talking about Nvidia. But what is a GPU? And who invented it?
I can answer the first question for you. At least in enough detail for you to understand how a GPU differs from a CPU. But as for who invented it, the answer is nobody. It slowly evolved over the course of the development of computers and the need to power their displays.
Let’s help you Know a Little More about GPUs.
GPU stands for Graphics Processing Unit. You likely think of it as the thing that makes your video look better, particularly for gaming. Depending on your profession you might think of it as the thing that speeds up your editing or rendering. Crypto miners have often used them to process calculations needed for minting crypto coins. And if you work with the many AI models out there, you probably rely on it for some of your model processing. But what is it exactly?
At base, a GPU can perform certain mathematical calculations at high speed. That’s not that different from a CPU so what’s the difference? The GPU is designed to be able to perform the same operation on multiple data values at the same time. A GPU doesn’t need to do all kinds of operations like a CPU does. That gives it the ability to push through the instructions it is executing at a higher speed.
The earliest GPUs were designed specifically for controlling the display. Over time they have become more programmable to execute parallel-processing for multiple kinds of tasks.
And the GPU is different from the graphics card. The card is a combination of the GPU some memory like VRAM, ports like HDMI and cooling. The GPU can be included on things other than graphics cards, like integrated on the same card as the CPU.
The modern GPU contains a number of microprocessors, called cores. CPUs can have multiple cores, usually around a dozen or so. GPUs have hundreds. Each multiprocessor has a shared memory block and registers. The overall GPU has its own memory plus device memory from the board it’s attached to.
You may have a machine with what’s called an integrated GPU. This means you don’t have graphics card, just the GPU. The GPU processors are on the same card as the CPU for greater efficiency, though usually with a sacrifice of performance. You can add a discrete graphics card to a system with an integrated GPU and then decide in the OS which one to use.
A discrete GPU is the traditional GPU, that operates separately from the CPU. Discrete GPUs that handle graphics usually have their own graphics card. For other tasks the GPU may be on a different kind of card – like a special purpose one for cryptocurrency or AI purposes, or slot right into the motherboard.
So how did we get from just a CPU to needing a GPU?
Before the GPU there was the graphic controller. It let the CPU do most of the processing, and just coordinated what lights on the display to turn on and off. But it was too hard to do 3D with that kind of system. Work on 3D imaging experimented with single processors controlling single pixels in order to combine pixels in a short amount of time and render graphics that appeared 3D. Out of this effort came the GPU.
Arcade systems were the earliest examples. They used specialized video chips that composted data as the display was sent to the monitor. These became more sophisticated as time went on. Namco’s Galaxian used special graphics hardware in 1979 that supported RGB color, multi-colored graphics sprites and tile map backgrounds. Several other arcade companies besides Namco started using the Galaxian hardware.
The first implementation of a graphics display processor as its own integrated circuit chip came from NEC in the 1980s. It supported 1024 x 1024 resolution and was used on multiple graphics cards including the Intel’s first GPU, the 82720.
A few other notable GPUs came out before we got to the first modern discrete GPUs in the 1990s.
Hitachi had the first CMOS graphics processor for PCs with up to 4K resolution in monochrome,
The 1985 Commodore Amiga had a custom graphics chip and a coprocessor capable of manipulating graphics hardware registers. Texas Instruments released the first fully programmable graphics processor in 1986, the TMS34010. That eventually became the TIGA Windows accelerator card.
But a watershed came in 1987. IBM released its 8514 graphics system for PC compatibles. It also introduced Video Graphics Array, aka VGA with a max resolution of 640 x 480. NEC followed in 1988 with the creation of the Video Electronics Standards Association or VESA which developed Super VGA or S-VGA with up to 800 x 600 pixels.
And in 1991 S3 Graphics introduced the S3 86C911 which added 2D acceleration support. 2D accelerators quickly pushed general-purpose graphics coprocessors out of the market.
Sony coined the term Graphics Processing Unit, or GPU in 1994, referring to the 32-bit graphics processor from Toshiba in the original PlayStation.
Arcade systems began developing 3D graphics systems and there were several attempts in the mid-1990s to bring that workstations and home computers. Fujitsu took some of its work for Sega’s Model 2 arcade system into a single integrated circuit for home computers in 1995. The Fujitsu Pinolite, the first 3D processor for home computers was released in 1997. Mitsubishi released the first GPU capable of transformation and lighting for workstations in 1997.
Home computers usually had access to 3D graphics through 2D accelerators that tried to add on separate 3D functions, like the S3 ViRGE , ATI Rage and Matrox Mystique. True 3D graphics were possible with a discrete board but those lacked a 2D interface, such as Power VR and 3dfx Voodoo.
Renditions Verite chipsets were among the first to integrate 2d acceleration and 3D functions on one chip.
But the first of what we would consider modern GPUs in the 1990s combined the graphics controller with a rendering engine and lighting engine on a programmable chip that could work independent of the CPU. That off-loaded processing which sped up the CPU and thus the entire machine when graphics-intensive work was happening. It improved the performance of design programs, like AutoCAD as well as video game graphics rendering.
Fujitsu developed one called the “Thriller Conspiracy” project but never made it to market. Nvidia was the first to bring it to consumers with the GeForce 256 GPU also known as the NV10 in 1999. Nvidia called it the “world’s first graphics processing unit.”
Nvidia continued to lead from there, producing the first chip capable of programmable shading, called the GeForce 3, used in the original Xbox.
In 2007, Nvida released CUDA, a software layer that made parallel processing generally available on the GPU. That meant you could do other things besides just the graphics the GPU was designed for. Eventually that led to alternative uses like cryptocurrency mining and deep learning model processing among others. Anything that required a large amount of processing power with a narrowly-defined outcome.
There’s lots more to GPUs including Ray-tracing video acceleration and more. But this should give you a good sense of what a GPU actually is and what makes it different from a CPU. In other words, I hope you know a little more about GPUs.