Inspired by a very short discussion on Twitter, I’m now curiuos about Nvidia history, especially the early years. There is a very brief company history on their web site, as well as some very short informations on Wikipedia. Besides that: is anybody of you aware about more material on founding and early years of Nvidia? I’d take anything from websites to books, video, audio, newspaper articles and such. I didn’t do extensive research on my own so far.
Depending on the material I find, this might end in some VCFe/VCFB presentation and/or in a Youtube video or some sort of podcast.
(I can’t read Twitter from here, so I don’t know what was said in that discussion).
It’s not mentioned on the Nvidia wikipedia page, but I definitely remember that a bunch of SGI graphics engineers moved over to Nvidia around the time of Belluzzo’s CEO period (yet another former MS exec - I guess you know what later happens to companies when a former MS exec takes over)
This story may be related: https://www.eetimes.com/sgi-graphics-team-moves-to-nvidia/
I worked for SGS-Thomson, later known as STMicroelectronics, and informally as STM or ST. (This was something of a mistake, as I thought I’d joined Inmos, but the transfer of ownership had already happened.) I’ll call it Inmos… Within Inmos in Bristol, there were two groups, the transputer group and the graphics group. Both had been designing chips using in-house tools and transputer-based workstations, but gradually both groups moved to using HDLs, and modern commercial tools - in the case of the transputer group, where I was, that was synthesis with Synopsys, and physical design (layout) using Cadence tools, and design rule checking with Dracula. Not sure whose simulation tools we used - possibly Mentor’s. There was a lot of innovation followed by acquisition and assimilation in the EDA tooling space.
Anyhow, my recollection is that there was some collaboration between the graphics group and NVIDIA, and some placement of engineers, and transfer of knowledge. Whatever the product of that, the upshot was that NVIDIA’s next project was pursued without ST.
My takeaway was that ‘we’ taught ‘them’ how to design and build chips.
In 1993, [Peter McGuinness] was instrumental in setting up the SGS-Thomson/nVidia joint development partnership. The strategic vision behind this project was to combine the emerging multimedia PC with the new possibility of integrating 3D graphics into a single chip to transform the PC into a fully fledged entertainment platform.
I note that '93 is also given as NV’s founding date.
My informal understanding is that NV won the GPU wars by relentlessly shipping new and improved product twice a year, for the trade shows, and any company which missed one of those self-imposed deadlines lost market share.
The Twitter thing is in German anyway. I tweeted a short article about the beginnings of Sun ZFS. A friend of mine commented, that there was a “ragequit” by the Solstice team (Solstice Volume Manager), which added to the beginnings of ZFS as well and brought as another example, that something similar happened to Nvidia with a SGI team.
I found the EETimes article as well, but this reads more like a partnership and not as a team of unhappy engineers left SGI in order to work for Nvidia. Also, the article is from 1999. At this point, Nvidia was already 6-7 years old.
You might find pointers to further documents by searching The Register. For example, I found this:
Interestingly, Nvidia also helped Microsoft in the development of the DirectX 6.0 3D graphics component Direct3D. Its contribution: adding multi-texturing to the games-oriented API. As Direct3D is designed to provide a common link between games software and graphics acceleration cards, to an extent it dictates how the cards work. In other words, any card that supports Direct3D 6.0’s multi-texturing must ultimately be based on Nvidia multi-texturing technology. Which, allegedly, it copied from 3Dfx. So, support multi-texturing through Direct3D 6.0 and your card is in danger of infringing 3Dfx’s patents too. Microsoft is already recommending Nvidia’s Riva TNT as the DirectX 6.0 graphics reference platform.
I see NV started with an ST co-development and used ST fabs (ST was very much a massive manufacturing outfit, could be viewed as designing chips in order to fill their fabs, rather than in order to sell chips.) And I see NV moved to TSMC after Riva generation, for the TNT generation. Some useful stats and crosslinks on this GPU site. (I believe NV stayed with TSMC ever since then, although I could be wrong. I don’t see them mentioned very prominently in this document about the phenomenon of the fabless semiconductor company, and that’s a slight surprise to me.)