dave please stop your 3dfx pr crap!
kristof 3dfx:
I keep hearing this and I can think of a few incorrect decisions but they don't seem to fit in with this discussion topic. So what wrong decisions did 3dfx make and how could they know in advance that they were bad ?
One of the bad moves I know is that they put way too many eggs in one basket, actually to some degree you could say that NVIDIA made the same mistake. 3dfx and NVIDIA (at the GF1 time) were putting everything on one single product : VSA-100 and GF1. Neither company had fallback products. 3dfx is now, be it late, fanning out into other things like the TV Tuner. NVIDIA is now also fanning out with Motherboards, sound, etc... But essentially NVIDIA is still building on its "one" core product (the base GF1 core). If the next core fails all the other products, except for the sound, still depend on it. So both need fallback and both now are making those moves, 3dfx in multimedia and NVIDIA in sound. Not doing this fanning out is what I call a mistake, and 3dfx has been suffering from this mistake "big" time. If GF1 had seen a glitch (whatever kind) then NVIDIA would have been hurting in a "very" similar way, since they had no real fallback either.
Now what other incorrect management decisions did they make ? Were the multi-chip, .25, and SDRam bad decisions ? Can we call them management decisions ? Point is IMHO this could have worked just as well as selecting single chip, .18 and DDR. The difference is that .25 and SDRam were certainties while .18 and DDR weren't. Naturally 3dfx failed to execute their idea, and this failure in execution has hurt them. But who do you blame for that ? What went "really" wrong in the execution ? Was the multi-chip idea to blame for it ? Was it an old tool they used ? Was there a bug in a library ? Did the fab f..k up ? Did they fail to secure enough production ? Did a single engineer fail to fix a bug ? Point is we'll probably never know. But if they had executed the .25, SDRAM and multichip - it "could" have been sound decisions.
So what did management really do wrong ? What are the factual mistakes they made ? Today its easy to say multi-chip was bad, but if you trace back to the start you have to admit that it could have worked just as well... at least thats what I think.
So please enlighten me on what went wrong... what did management do that was soo bad ?
Leaving the chip market is also arguable a bad move, they pretty much handed all deals to NVIDIA since ATI and Matrox don't sell to board makers. S3 left the market and Imagination hasn't been very successfull - so what could Creative Labs and others do ? But how did 3dfx reason about it... they saw their margins getting smaller and smaller, more companies more competition more different Voodoo2 products... in the end competition was soo huge it ate away the 3dfx profit too. So they reasoned: single source, we determine the price, we maintain the full profit margin. Seems like 3dfx allowed itself to be squased by Creative Labs while NVIDIA maintains high prices no matter how many deals they have. So did they screw up there, probably.
To me it seems like 3dfx is acting more and more like Videologic. Videologic has been suffering bad management decisions for a long time (at least according to its investors). Videologic also had huge costs (made a loss for many years in a row IIRC), they also make boards and they have nice technology but they fail to execute on time to compete. Videologic has re-structured into IP business and 3dfx seems to act in similar ways. 3dfx is re-structuring, they are talking about IP licensing. So they look quite similar to me.
Ah well... I am curious how the next gen of all companies will do... wonder where each company will end. People have lost faith in 3dfx and its understandable, we'll see what Spectre brings: the end or the re-birth. What do you people expect of Spectre to compete ? Does it have to beat NV20 in all playing fields : Performance, Features and Prize ? Or will one of those do ? For example ATI competes mainly with Features at the moment. KYRO competes a bit with Prize... where do you expect 3dfx to be - since right now they don't really compete in any field (prize, performance, features) ?
My pal mr poster:
quote:
--------------------------------------------------------------------------------
Originally posted by Kristof:
Care to elaborate on those misguided decisions ? What are they and Why are they misguided ?
--------------------------------------------------------------------------------
I was hoping that nobody would ask as now I have to spend lots of time composing a reply. I mentioned some of these already, but I try to explain myself this time.
Multichip scalability
This was the source of all the trouble. It misfocused the project as the there was no need to seek solutions to make individual chips fast (just throw in some more!).
They underestimated the complexity of interchip communication (at least partial reason to v5 lateness and v56k troubles) and probably the related driver issues .
The mainstream solution (v5) is double chip, which is horrible costwise, hence bad margins and high prices. For multichip to work financially, the mass market product should definitely be a singe chip.
Synchronized core/memory
This is related to above. 3dfx totally underestimated how fast high speed DDR would become cost effective. With their synchronized design, they were unable to take advantage of the high speed memory as it become available.
Nvidia was initially critisized for having a fast core that was seriously hampered by the memory bandwidth. Now they are selling about the same chip with huge premium as the Ultra. Great stuff businesswise.
0.25 process tech
This was supposed to be easy and well known and some 3dfx rep was actually ridiculing nvidia for choosing the more advanced tech for geforce. They had troubles getting their complex design to work with the old process (contributing to initial delays) and in the long run, it has become very ineffective cost wise.
T-buffer
Ok, no one will agree with me here, but I consider T-buffer a major hack. FSAA should not be done in the video out level. The hackiness actually shows up when trying to get screenshots and in lack of windowed FSAA. Please note that I'm NOT critisizing rotated grid, thats obviously the right way to go. This somewhat academic argument, businesswise what they did here was clearly the right thing.
Lack of high quality OpenGL drivers
There is clearly a decision to put engineering resources elsewhere. While gamers don't care, developers and businesses do. What game developers use in their dev machines today is what gamers have in their machines tomorrow. If 3dfx had bothered do this work early, it would now be much easier to develop drivers for Rampage and other upcoming products.
(apparent) Discontinuety from VSA to Rampage
Some people want to see revolutionary tech. However while less sexy, evolution is usually more effective and much easier to accomplist. Witness how far Nvidia has come by incrementally improving their product. I don't expect NV20 to be a revolutionary either.
3dfx has fallen so far behind that they must try revolutionary generation leap with rampage. Everything is supposed to be new. This means that they must deal with all the complexities at the same time instead one at time. I'm afraid this will overhelm them and cause delays and other problems.
Witness how long it took nvidia to understand how to effectively drive their T&L hardware. Now they understand the issues and can consentrate on other things.
Ok, I'll stop now. I think there were some other things I had in mind too, but I have already happily forgotten them.
quote:
--------------------------------------------------------------------------------
IMHO If things would have gone to plan (read no 6+ month delay for whatever reasons there have been) then things would have been different, 3dfx would have had low to high-end products. The 6 months delay set the window "wide" open for NVIDIA.
--------------------------------------------------------------------------------
I think in that case VSA might have been moderately successfull and 3dfx would be in better financial position now. Nvidia would still be dominating.
I believe that VSA's lateness was due to wrong technical decisions. It was not something that just magically happened.
quote:
--------------------------------------------------------------------------------
I am also not sure about your 2- and 3-way marketplace. Never underestimate the competition and there is more than just 3dfx and ATI. Matrox is still kicking, Imagination technologies also is, S3 remains in two of the markets NVIDIA wants to compete in (Laptops and Integrated). So 2-way is a bit over simplistic... definitely don't forget the famous BitBoys they have been beating NVIDIA for over a year now
--------------------------------------------------------------------------------
I think that the marketplace is currently two-way. ATI and NV are the only significant players. 3dfx (and everyone elses) volumes are too low to register in the radar. I hope that 3dfx can get back to the battle.
3dfx is getting attention (this post included) as much as it does only because of historical reasons.
Phew.
i suggest all nvidia and 3dfx supporter should read this!
One of the best threads over 3dfx and nvidia!