On October 11, 1999, Nvidia released its GeForce 256, its first product that it described as a GPU.
What made the Nvidia GeForce 256 the first GPU?
I’m sure 3dfx fans would argue that 3dfx made GPUs, and it seems a little arbitrary to say the GeForce was a GPU and the TNT2 wasn’t. It was more powerful than the TNT2 and offered more features, but what feature did the GeForce have that made it a GPU and the TNT2 not a GPU? The transform and lighting engine, maybe?
The concept of offloading work from the CPU onto a graphics coprocessor was nothing new. Commodore did it with the Amiga way back in 1985. And 3dfx was doing it on PCs as early as 1996.
But Nvidia didn’t use the term until October 1999. It continues to use it today, though, and it also uses the GeForce brand.
The GeForce 256 wasn’t necessarily an immediate game changer. It was a real boon for systems with low-end CPUs like 300 MHz Celerons, but faster CPUs could do transform and lighting in software faster than the GeForce 256 could do in hardware. But the concept of offloading these tasks from the CPU was a good one. A lot of gamers at the time were on a budget and had low-end CPUs like Celerons. Nvidia stayed ahead of Intel by creating follow-up graphics cards more frequently than they had in the past. Nvidia’s follow-up GeForce GPUs fared well.
The GeForce 256 wasn’t Nvidia’s first product and wasn’t their first graphics card, but if they want to say it was their first GPU, I’m not sure too many people will argue with them. Or it could be the GeForce 256 was the first just because Nvidia didn’t invent the term until the GeForce 256.
Nvidia bailout by Sega
Today Nvidia is one of the most successful companies in the world. But it struggled early on.
Three years after its founding in 1993, Nvidia partnered with Sega to design the video for the Dreamcast console in 1996. Nvidia’s chip wasn’t up to the task Sega was asking of it, they were out of money, and couldn’t finish the project. Sega bailed them out with a $5 million investment in the company. Sega made its money back along with a $10 million profit, so it worked out fine for Sega in the end. The Dreamcast ended up with a chip manufactured by NEC in the Dreamcast, after Sega also considered a chip designed by 3dfx.
The money from Sega helped Nvidia introduce the Riva 128 chip, which sold more than a million units and led to the successful TNT2, followed by the GeForce 256. If it hadn’t been for Sega’s investment, Nvidia could have gone out of business before 3dfx did, and perhaps 3dfx would have survived instead. Instead, Nvidia purchased most of 3dfx’s intellectual property on December 15, 2000. 3dfx wound down operations and went bankrupt October 15, 2002.
Thank you
This blog post got a lot of attention. Thank you to those of you who shared it, and thank you for reading this far. I post retro content every weekday, so if you enjoyed this or found it informative, I hope you’ll check out some of my other recent content, and maybe come back again soon.
David Farquhar is a computer security professional, entrepreneur, and author. He has written professionally about computers since 1991, so he was writing about retro computers when they were still new. He has been working in IT professionally since 1994 and has specialized in vulnerability management since 2013. He holds Security+ and CISSP certifications. Today he blogs five times a week, mostly about retro computers and retro gaming covering the time period from 1975 to 2000.
I miss the days when NVIDIA cared solely about its customers and products.
Today, it has turned into an AI company and is showing increasing disregard for its consumers with each passing day.
What is happening with the RTX 5090 is outrageous.
Time will tell if this behavior will bite them in the ass or if it’s the right move.
The certainly make a ton of money now. But Comodore did the same with their home computer “wars” but it bit them after (beside other things).
That’s a good analogy. I’ve written a little about Commodore here, and how more than one of its short-sighted decisions cost them in the long run.
FYI, “T&L” stood for transform & lighting, not texture & lighting. All the early cards did texturing as that made the biggest difference in graphics quality.
Thank you, I will make the correction!
Atari did it in 1979 with the Atari 800. Its graphic chip ANTIC is a full bus mastering cpu with its own instruction set and is an equal to the 6502. Neither needs the other to do its thing, they literally just take turns being in charge of the bus.
Ah yes, good point, I forgot that was one of the many cues the Amiga took from the Atari 800.
According to Wikipaedia the term GPU has been used since the mid-’70s. As you noted there were plenty of graphics cards which could easily be described as GPUs, and (also as you noted) for NVidia to start calling an arbitrary card a GPU doesn’t mean anything.
Also interesting back in those days (late-90s-early-2000s) it was common (for hard-core gamers, at least) to have a separate graphics card and 3D card (i.e. GPU).