It's more like thanks to 3dfx?

dagw 8 days ago | parent | next [-]

3dfx never really competed with SGI since they were never compatible with the commercial and scientific software that people bought SGI machines for. Nvidia on the other hand (mostly) was. I worked at a small animation studio at the time and shortly after the GeForce 2 was launched we'd basically replaced all our expensive SGI and Intergraph machines with cheap generic Wintel boxes at a quarter of the price.

▲

FirmwareBurner 9 days ago | parent | prev | next [-]

There were like a bazillion companies competing for consumer 3D accelerators in the 90s. 3dfx was the most successful thanks to their Glide API and vertical integration but they weren't the only one on the market, which is why they were so affordable despite the novelty. Unlike today.

▲

pjmlp 8 days ago | parent | prev [-]

Not really, because initially they went with Glide, their great boards eventually weren't a match to Nvidia, that had enough cash to buy 3dfx.

I was disappointed that I couldn't make my newly bought Voodoo card work on my motherboard due to a PCI connection issue, but the Riva TNT that the shop offered me as possible alternative did, thus NVidia got one more customer.

▲

flohofwoe 8 days ago | parent [-]

I think what parent means is that 3dfx was founded by three former Silicon Graphics engineers, so I guess that the 3dfx hardware had a lot more SGI DNA than Nvidia's chips.

Nvidia's first 3D chip ~~Riva 128~~ (my bad, it was the 'NV1') also was a weird design, and it's successor the Riva 128 wasn't remarkable performance-wise especially compared to what 3dfx had to offer (Nvidia's only good decision was that they bet on D3D early on when everybody else was still doing their own 3D APIs - even though early D3D versions objectively sucked compared to Glide or even OpenGL, it turned out to be the right long-term decision).

Nvidias first remarkable chip was the Riva TNT which came out in 1998 (hardware progress really was unbelievably fast back then - 3dfx Voodoo in 1996, Riva 128 in 1997, Riva TNT in 1998, and both the Riva TNT2 and Geforce in 1999).

edit: fixed my NV1 vs Riva 128 mistake, somehow I merged those two into one :)

	▲	ChrisGreenHeur 8 days ago \| parent \| next [-]
		A better point is that the consumer 3d cards were created specifically because SGI was designing 3d machines the wrong way. It became obvious that the way to design a consumer 3d card was to create a small multithreaded chip that you could scale up and down based on the workload. SGI instead created special designs for every computer, sometimes multiple graphics board designs for every computer. Sometimes (such as VICE in an O2) SGI would also put old processors in odd configurations inside new computers in the hopes they could offload things such as dvd decompression, but then drop dvd support before releasing the machine while keeping the hardware. SGI was just very unfocused around 1998-2204 when this whole consumer 3d chip became realistic and they just refused to do things in a sane way. they even knew it but did it anyway. Betting the company on web servers instead.
	▲	pjmlp 8 days ago \| parent \| prev [-]
		Point taken, and as usual there is a lesson learnt with first movers losing the market they helped create, 3dfx on one side, and Ad Lib for the other side on multimedia PC history.