Designing new chips takes a lot of time and manufacturing chips takes quite a bit of time as well. The VSA-100 chip was supposed to be ready sooner than it was, it was going to compete with the GeForce 256. As far as I can tell, acquiring STB and getting into the business of selling their own cards, sidetracked 3Dfx in a lot of ways, which means the Voodoo 5 got a bit delayed and as such wasn't very competitive when it came out. Delays and mismanagement also caused their Rampage project to stall, that one was supposed to have a T&L unit as far as I recall, but it was going to be another chip on the board(?), which always sounded to be like something was definitely not going well over there.

Garrett W wrote on 2024-04-10, 12:56:

Designing new chips takes a lot of time and manufacturing chips takes quite a bit of time as well. The VSA-100 chip was supposed to be ready sooner than it was, it was going to compete with the GeForce 256. As far as I can tell, acquiring STB and getting into the business of selling their own cards, sidetracked 3Dfx in a lot of ways, which means the Voodoo 5 got a bit delayed and as such wasn't very competitive when it came out. Delays and mismanagement also caused their Rampage project to stall, that one was supposed to have a T&L unit as far as I recall, but it was going to be another chip on the board(?), which always sounded to be like something was definitely not going well over there.

Mostly true there. The Spectre line was setup with a separate geometry chip (I vaguely remember Sage as the codename) with everything else on a separate chip (Rampage). The entry level line was 1 Sage and 1 Rampage and the high end was 1 Sage and 2 Rampage chips. From the benchmarks that have come out using ES hardware and pre-alpha drivers it was absolutely going to be a monster if it made it to market.

They probably didn't think the CPU was a better road against the T&L, but it was clear already since the Avenger chip into the Voodoo3 that it was already too late and the game over, with the late VSA-100 feeling just what a 1998 "Voodoo3" should have been and probably would have failed anyway without a very soon next generation chip. The Voodoo Graphic and for a few time the Voodoo II / SLI solution is what they should be remembered for imho.

While TnL was a natural and useful evolution for graphics chips, it wasn't easy to argue for it back then. The systems were balanced and expected to do it with a CPU and those were also speeding up nicely.

It wasnt so much the case of 3dfx being slow to innovate, it was Nvidia murder spree.
3dfx guys were going at steady SGI small steps pace of a 3D workstation company. Nvidia came at it with LSI Logic attitude (CEO came from AMD thru LSI) of quickly iterating and shipping silicon.

My favorite obligatory 3dfx fall/Nvidia killer mode timeline Re: Bought these (retro) hardware today

Riva 128 (April 1997) to TNT (June 15, 1998) 14 months
--Revenue 1997 Nvidia $29M, 3Dfx $44M
--Revenue 1998 Nvidia $158M, 3Dfx $202M
TNT2 (March 15, 1999) 8 month
GF256 (October 11, 1999) 7 months
--Revenue 1999 Nvidia $375M, 3Dfx $360M
GF2 (April 26, 2000) 6 months
3dfx dies here
--Revenue 2000 Nvidia $735M, 3Dfx <$200M and bankruptcy
GF3 (February 27, 2001) 9 months
--Revenue 2001 Nvidia $1.4B
GF4 (February 6, 2002) 12 months
FX (March 2003) 13 months

Lets not talk about NV1/NV2 duds 😀 After hail mary Riva 128, with tringles this time, Nvidia started working on at least two designs in parallel. One to ship as fast as possible, the other implementing good but costly ideas. At the time first one was shipping the next ones costly ideas were already cheap and so new even more ambitious project was created and so on. Nvidia also employed state of the art logic simulations - all designs were emulated on a gate level before tapeout. 3Dfx was also emulating, but on a subsystem level, Low-level emulation (LLE) vs High-level emulation (HLE) situation.

ATI was kept alive for the most of this time with B2B deals, bundling pitiful Rage chips in name brand computers.
edit: updated with more financial details

Sure nvidia timings were impressive for most companies but there was no coming back for 3dfx after releasing of both the Avenger and the weak VSA-100 that late and that obsolete at 0,25um when basically any other chips were technically superior; to take time in such market and such period was not possible for 3dfx. ATi did their best to survive while 3dfx idea to sell own cards was incredible, expensive and mostly as late as possible. 3dfx did too many errors since the Rush and none would be able to solve such complicated situation.

Avenger was solid and traded blows with TNT2, made them lots of money. It was definitely the VSA fiasco that killed 3Dfx.
I had V3 3000 at the time, my friend had V4 4500, there was miniscule fps difference in UT99 😮
Who decided releasing another DX6 chip a year after previous DX6 chip was a good idea? :0

October 2000 prices:
$150 voodoo 4 4500at CompUSA
$150 low end geforce2mx murdering it easily in all games except glide
$70 Voodoo 3 3000

rasz_pl wrote on 2024-04-11, 12:46:

Avenger was solid and traded blows with TNT2, made them lots of money. It was definitely the VSA fiasco that killed 3Dfx.
I had V3 3000 at the time, my friend had V4 4500, there was miniscule fps difference in UT99 😮 Who decided releasing another DX6 chip a year after previous DX6 chip was a good idea? :0

Exactly, they should have made a choice between Avenger and VSA-100. They were both fabricated on the same node as well so the release timeline wouldn't have been that impacted one way or the other. You certainly can imagine a scenario where 3DFX cut the fat a bit even further (don't bother with Rush, the STB acquisition, and prioritize VSA-100 over Avenger) to stay on top of things and make it to Rampage. One way or another I think we would have still only ended up with 2 major vendors, but there's a chance it was 3DFX buying up ATI in that scenario rather than AMD (which would have had wild implications, like NV buying out AMD in their post Conroe lean period).

3dfx bought STB factory to penetrate OEM market. They probably cut some expenses on RnD and it all went downhill from here. Avenger was supposed to be low-end refresh of Voodoo Banshee , but something gone wrong so they had to use it as full-fledged Voodoo 3 and bullshit their way through missing modern features. VSA-100 was supposed to be the true competitor to TNT2/G400/Riva 128 Pro.

Avenger was solid and traded blows with TNT2, made them lots of money. It was definitely the VSA fiasco that killed 3Dfx.

Avenger was a horrible lineup that started the downward spiral, because it had absolutely no future proofing.

Neither Banshee nor Voodoo3 were low-end, they performed too well. Come to think of it, Voodoo3 could benefit from TnL the most, as there weren't CPUs fast enough to carry her to the top at launch. But it wouldn't be economical.
3dfx fans often say this or that should have been like the next chip, but that is wishful thinking rather than correct (and possible!) business decision.

OEMs come to you when you have the right product. No need to buy factories.
IMO the problem of 3dfx was inability to repeat the success of Voodoo3, alienation of card vendors by competing with them, lack of low-end parts, and only tertiary general obsolescence which included TnL.

I had a just released Voodoo3 2000 for the new K6-2 I bought but it already was like something wasn't working in that vga serie even with all the success of the early cards just behind. It was clear for a consumer that something went wrong and with such high price.

Then I am sure the T&L as often said wasn't really needeed at the beginning but able to attract investments and research and the whole Geforce was anyway quite a good gpu beside that. 3dfx needed some heavy reasons for the consumers to go for their cards.

386SX wrote on 2024-04-11, 14:38:

I had a just released Voodoo3 2000 for the new K6-2 I bought but it already was like something wasn't working

spoiler: 3dfx wasnt your main problem 😀

rasz_pl wrote on 2024-04-11, 15:56:

386SX wrote on 2024-04-11, 14:38:

I had a just released Voodoo3 2000 for the new K6-2 I bought but it already was like something wasn't working

spoiler: 3dfx wasnt your main problem 😀

I know the CPU and mainboard were slow even when new but still a low to middle-end config (at least in Win 98) at that time in my country, bought brand new when most people I knew had Pentium-MMX and Pentium configs. It wasn't necessary the card speed because I played low resolution early accelerated games anyway which ran great but review after review it was clear 3dfx didn't get it right like nowdays is obvious and already suspected in most reviews. I didn't care about 32bit colors or high resolution CRT gaming. It was more about the feeling the money I spent on the brand new card at the store were not well spent anyway.

There is a saying where I live that perhaps the translation is not correct, "you can't live off past glories".
I think that at 3dfx they wanted to live the entire life of their first success; I've read some articles that talked about parties and lack of control, and uncontrolled spending of money. I don't know if it's true, but it seems possible to me.
I'm not a fan of 3dfx, but it was an important loss for users that helped a lot to get to the current situation.

Hoping wrote on 2024-04-11, 16:36:

I think that at 3dfx they wanted to live the entire life of their first success;

Both Voodoo1 and Voodoo2 render all of the triangles you throw at them all the way thru, and just at the very end of the pipeline check if said triangle passes Z-buffer check. You can feed geometry back to front, front to back, doesnt matter same speed. Unsophisticated brute force. I have no proof one way or the other, but I have a suspicion both Banshee and Voodoo3 do the same - that both are just miniaturized V2 architecture running with faster clocks, modified memory controllers and no other optimizations 🙁 Basically from 1997 to 2000 on same arch with zero advances other than Moore's Law.

Hoping wrote on 2024-04-11, 16:36:

I've read some articles that talked about parties and lack of control, and uncontrolled spending of money. I don't know if it's true, but it seems possible to me.

Meh, thats mostly BS. There is no cheaper way to bump morale and limit employee turnover than free lunches/drinks and paying for occasional party. Its literally peanuts compared to employee compensation.

DrAnthony wrote on 2024-04-10, 14:03:

Mostly true there. The Spectre line was setup with a separate geometry chip (I vaguely remember Sage as the codename) with everything else on a separate chip (Rampage). The entry level line was 1 Sage and 1 Rampage and the high end was 1 Sage and 2 Rampage chips. From the benchmarks that have come out using ES hardware and pre-alpha drivers it was absolutely going to be a monster if it made it to market.

Perhaps you have different information, but from what I know the Rampage ES boards did barely run and drivers were horrible too.

Also I think the slowest Rampage was meant to be just the single Rampage chip - so no TnL and no Vertex Shader.

Anyway, considering the basic hardware specs, these cards wouldn't be that impressive in my opinion. 200-250 MHz 4 ROPS / 4 TMU chip made at 180nm process. I think we estimated many years ago the middle variant of 1x Rampage + 1x Sage could be around Radeon 9000 performance at best. That is just barely above GeForce2 Ultra and much slower than GF3. The 2+1 variant could be compatable to GF3 or Ti500 maybe, but at cost of much higher power consumption and huge PCB.

Also considering the first Rampage samples were barely working in november 2000, the cards could be ready for market in mid-2001 perhaps. So too late for any significant success.

The final VSA-100 products from PowerColor were marketed as compatible solutions with good value, as opposed to fast GeForces with newest features.

rasz_pl wrote on 2024-04-11, 17:30:

Meh, that's mostly BS. There is no cheaper way to bump morale and limit employee turnover than free lunches/drinks and paying for occasional party. It's literally peanuts compared to employee compensation.

I agree, there is no problem as long as you have the money for those parties, dinners and meals, the problem is that 3dfx was losing money and going to parties, dinners, and meals was not going to help them develop a chip with support for T&L, but it helped them spend the money that was already scarce, this can be seen by the profit figures that another user has posted above.
Poor management, the reason why many companies have closed. I don't understand much about business management, but I do know that spending more money than you earn does not give good results.
A shame, I think that the staff that had 3dfx would have been able to continue changing the game.
I also think it was a mistake to hold on to Glide and not see that Microsoft was bigger than them and was pushing D3D very hard with the help of others.
It seems that in 3dfx they believed that they could fight the entire world alone.

Hoping wrote on 2024-04-11, 17:58:

I don't understand much about business management, but I do know that spending more money than you earn does not give good results.

like Amazon? :--) https://www.forbes.com/sites/jonmarkman/2017/ … its-no-problem/ https://247wallst.com/special-report/2022/06/ … -turn-a-profit/
When you are on stock exchange profit is not important, market share and growth are.

3Dfx failed to grow in 2000. They didnt survive Nvidia relentless R&D assault and lost market share to Nvidias low end offerings. It wasnt GF256 that killed 3Dfx, nor was it GF2 or GF3. It was TNT2M64 and Vantas flooding market in huge volume at ridiculously good prices.
I dont recall a year 2000 game that would struggle on V3 3000? but people saw Geforce benchmarks and picked winning brand.