Anand Lal Shimpi's new Video Article - But WHY did 3DFX lose the technology race?

[Bojo]

I just read Anand's new article "Looking Back at 2000 Part 2: The Graphics Industry" what Anand doesn't address is exactly why 3DFX continued to release such crap graphics cards.


Basically as far as I can tell 3DFX's real problems started when NVIDA released the GeForce chip. At the time everyone new that it was some pretty hot stuff but 3DFX just kept on with the same old Vodoo chips.

Why didn't 3DFX develope a chip that was faster than the GeForce? Too costly? It seems bizaare to me that such a giant in the graphics industry simply stopped R&D on making kickass chips and instead focussed on making cards with 2 and 4 chips! Looking back it just seems ridiculous.

I know that my knowledge on this subject is not exactly very deep so does someone who knows more than I know why 3DFX stopped developing better chips?

 

[BenSkywalker]

They didn't stop, they just couldn't keep up with the torrid pace of nVidia.

If 3dfx had lasted another six months we would have seen their Rampage part, one that should have competed very well with nVidia's NV20.

If 3dfx had laster another year we would have seen their chip based on Gigapixel technology which reportedly would compete very well with nVidia's NV30 technology, but would have shipped six months earlier.

It came down to execution. I don't think you should think poorly of 3dfx for not being able to keep up with nVidia, as of yet noone in the industry has been able to. If you look at the bigger picture, ATi has lost a lot more ground to nV then 3dfx did, they just had a lot more ground that they could afford to lose.

 

[Sunner]

3dfx was never a giant.
In the beginning they were a giant in a very small market, when they died, they were simply one of three major players in a market that was definately bigger than in the beginning but still very small.

3dfx's troubles began when the VSA-100 was delayed, then delayed again.

Their real problem was that they didnt have any good OEM products.
Even with 90% of the retail market, they would still have been way behind nVidia and ATi.

And 3dfx didnt simply "stop", they focused on the wrong market, and even in that market, they couldnt quite keep up.

[edit]
Oups, Ben beat me to it this time, still puts me in a 1 point lead though
[/edit]

 

[Jethro Bodine]

It all started when Brian Burke brought back 3 bales of Mauii Wowee early last year.
By the time the 3dfx staff smoked their way through it/woke up, they were far behind in R/D.

"Dude, where's my market?!"

 

[merlocka]

It's a very fast paced industry with very short cycle times. nVidia's CEO has a quote something like "we are only 30 days from being out of business". nVidia executes like no other company, and 3dfx never got the OEM deals they needed. Man cannot live on retail alone.

 

[SuperSix]

Not to mention the Juarez Mexico plant that had MUCH higher labor costs than most other video card manufacturing plants, and that nVidia doesn't need to manufacture BOARDS in large volume. There's a reason STB was vulnerable when 3Dfx bought them, and the Juarez plant was one reason.

 

[jpprod]

From what I've heard, Matrox shouldn't be counted out of the game as the article seems to hint. I really shouldn't be saying this, but with no NDAs to hold me back, here it goes... I'm involved in developement of a DX8 game, and our lead 3D graphics guru had this to say about email discussions with folks at Matrox (they sent him a free G450 dualhead a while back BTW - that's what I call support )...

(translated from Finnish by me)
I got a email message from Matrox today (12/22/00) in which they confirmed me that a completely new card is coming soon. I was even given permission to ask questions about it. Message strongly implied that this time Matrox is going to seriously compete in speed depatment as well.

Matrox's always kept a very low profile when it comes to upcoming products. In my book, them approaching developers like this means that the G800 might not be that far off!

 

[Deeko]

In my opinion, it was the delay of the Voodoo5 that killed them. The Voodoo5 was intended to compete with the GeForce SDR, which it would have demolished. It even would have defeated the not-yet-released GeForce DDR. There was no MX at the time, so the Voodoo4 would have owned the budget market, this gaining 3dfx much needed OEM deals. The Voodoo5 is truly an amazing product. The fact that it is competing with a generation ahead of what it was intended, and still keeps up, is amazing. It's really a shame it was delayed so much, I think they would be on top of the market if it had been released when it was originally slated to be.

 

[pidge]

Matrox doesn't need to be in the speed race that badly. They have excellent 2D. Great drivers. They have a sector which ATI can't touch. They should just focus on that (the workstation market). The gaming market is just to competitive right now. Just keep their current OEM's happy. I think if they do that, they will be just fine.

 

[Mem]

<< The Voodoo5 is truly an amazing product >>



I don`t think it`s ,anyway 3DFX should of stayed to making chipsets this was when they were at there best &amp; strongest.They should not have brought STB &amp; started making their own cards,plus long delays between models which were nothing great when they finally came out. There is a lesson here for all companies &amp; 3dfx found out the hardway.

 

[sandorski]

Buying STB killed 3dfx. The Voodoo 4/5, if released on time, would have been more successful, but the delays certainly killed 3dfx not the technology.

I too will second that the Voodoo5 is an awesome card.

 

[OneOfTheseDays]

3dfx just made some STUPID ass moves. I mean if you want to get some oem deals then make your frickin product truly AGP compliant. They didn't support agp texturing, or whatever its called, which makes it impossible to integrate into the mb which is what many manufacturers are doing. They also pissed off creative, hercules, and other companies with the whole STB deal. The execs in charge of 3dfx just made some incredibly bonehead moves.

 

[BFG10K]

I partially blame 3dfx's foray into the Mac market for their death.

If 3dfx hadn't wasted time and resources on Mac drivers, 555 and bi-endian support on the VSA-100 and Mac technical support/PR, we might well have seen the V5 come out earlier and 3dfx have more money in their pockets.

Just what is it 3dfx hoped to achieve by supporting ~3% of the computing market? Macs are far too underpowered in 3D games to benefit from a decent video card anyway. Given ATi has a totally monopoly on OEM Macs and Apple provides no BTO options, most Mac users have to purchase a Voodoo (more $$$ in addition to their already expensive systems) and then simply throw out the ATi card which they've already paid for.

3dfx's support of Apple was a totally stupid idea in every conceivable way.

 

[DaveB3D]

Two words: bad management..

And Ben, you are wrong. Gigapixel would have competed (likely beaten the doors off of) NV30, but it would have been out around NV25.

 

[Mr.IncrediblyBored]

According to some anonymous poster at B3d (It has been partially confirmed he worked at 3dfx), Fear was set to tape out in early spring.

 

[Jethro Bodine]

So Dave, as an engineer with video card experience, are you going to nVidia?

Your thoughts on whether nVidia will use the Gigapixel tech?

Mostly- WHY WHY WHY did you guys not kick the V5 out the door Dec.1999, when it would have been a competitive product?!

 

[BenSkywalker]

Dave-

&quot;And Ben, you are wrong. Gigapixel would have competed (likely beaten the doors off of) NV30, but it would have been out around NV25.&quot;

My unedited quote-

&quot;If 3dfx had laster another year we would have seen their chip based on Gigapixel technology which reportedly would compete very well with nVidia's NV30 technology, but would have shipped six months earlier.&quot;

I said what you said Dave, how is that wrong????? Six months before the NV30 would be NV25(assuming they stay in step this time

 

[Lord Gwynz]

The Banshee and Voodoo3 is where I started seeing things go downhill for 3dfx. The Banshee wasn't faster than V2/SLI and the V3 wasn't worth upgrading to from SLI. I think 3dfx should've stuck with the 3D add-on card design for a bit longer, at least until they were ready to bring out the Voodoo5. Then there was the TNT's ability to render in 32-bit color, which was a pretty useless feature at the time but it was hyped up to death.

 

[merlocka]

Gigapixel would have competed (likely beaten the doors off of) NV30, but it would have been out around NV25.

Coulda-shoulda-woulda is another reason they lost the technology race.

Banshee woulda beaten TnT if it had 2 TMU
Voodoo3 woulda beaten TnT2 if it had 32bit and AGP texturing for marketing and OEM's

and in the above quote, replace Gigapixel with Voodoo5 , NV30 with NV20 and NV25 with NV15... and then realize that didn't happen.

I wish 3dfx was still in the gambit, but the industry became very compeditive and DDDfx management sat and watched. The HW/SW engineering was good, but no product execution.

 

[DaveB3D]

Ah, sorry Ben.. that is what I get for not reading more carefully.

Me.. I doubt I'll go to NVIDIA. I don't particularly care to, honestly.

NV and Gigapixel. They'd be smart to use it. Hell, they'd be smart to finish our next chip and use it instead of what they are developing. But I think they are to arragent to to do that (most companies would be, really) and it just likely isn't going to happen. I can't wait to see what NV30 and NV35 is. I'll really get a kick if they don't match the performance until NV40.

 

[Castellan]

3dfx made two fatal erros. One was buying STB, the other was not liscensing GLide. Think about it. By buying STB, they pushed themselves into a market they had little to no experience in. Making chipsets is different from selling boards. Don't even want to think about the royalties they would have made from liscensing GLide either.

 

[DaveB3D]

Gotta love how people pull things out of their ass... Meaning people say things as fact when they have no idea what they are talking about.

 

[NicColt]

They got gready, very gready...

When the Voodoo 2 came out they owned the market share, every company wanted the oem chips to make cards. Business was good, business was growing. then.....

&quot;Power is good but ultimate power corrupts&quot;

Then they got gready, when they said &quot;hey why have a middleman when we can sell the cards ourselves&quot;. Big Mistake, they had fed these companies the oem chips, and now they took them away. All these companies now had a video card production and the marketing but no chips.... hhhmmm where to get the chips??? hey let's ask nVidia. well well well nVidia had lot's of chips, it was like the first date after a divorce, and here ya go all the chips you want and cheap too...... &quot;oh but we have bandwith&quot; yannnn. Now all of a sudden 3dFx's best allies were now their main competition. Hhhhmmm let me see there is one Voodoo 3 here and the POWER of 20 TnT2/U cards that are cheaper than the Voodoo 3 to choose from. Well D'UHHHH...

&quot;You keep your friends close and your enemies even closer&quot;

They created their own downfall and it was too late to recover, &quot;hey guys now we have eom chips&quot; how could they, it was like getting a f*cked and not even a goodbuy kiss. Now the bitch wanted back in. Yeeeaaa righttt....

 

[pen^2]

IMO, 3dfx drivers team must have been a big factor. it has only happened as of late that v5 were comparable to the speed of GTS. there is no telling v5 would have had such quality drivers even if it came out to compete with geforce SDR... Nvidia has drivers team like no other's, no matter how much u b!tch about leaked detonators.

 

[ElFenix]

NicColt... the voodoo3 was cheaper than the tnt2 or pretty much its entire life cycle... too bad that card didn't have 32 bit color, would have made a great OEM checkbox feature. really, 3dfx is dead because they never got oem sales, for whatever reason. perhaps because they started in the add-in business, perhaps checkbox features, whatever. OEMs care little for 3d performance, regular joe-blow cares little for benchmarks, only for checkbox features. and big companies want stable drivers and low cost parts. somehow, 3dfx was behind on checkbox features from 1999 (their only profitable quarters were after the v3 was released, BTW) and didn't get the v3 into OEMs.