Talk:3dfx/Archives/2012

Page contents not supported in other languages.
From Wikipedia, the free encyclopedia

General

I edited the Voodoo3 section with more info on the STB/3dfx merger. I was at 3dfx at the time and there is a lot of great info that never made it out. The STB merger single handily handed the graphics chip market to NVIDIA by ensuring OEM's would move to NVIDIA (STB quality problems were immense to the OEM's) and that then current 3dfx 3rd party manufacturing and branded partners would move to NVIDIA as the only choice to remain in business. The revenue 3dfx had from those 3rd parties was never replaced with the direct sell model.


@the company did not sell any Voodoo 4 or 5 chips to third party manufacturers Hope I'm doing this right, first time. The above statement from the Voodoo 3 part of the article is not true, there were 3rd party manufactures of Voodoo4/5 cards, albeit few and they came in late. I'm not an expert hence can't enumerate, but there was at least Powercolor with e.g. PowerColor EvilKing 4 4500.


A brief comment about this night's big edits in the chip table, I notice some of Geri's info being correct, and in some ways better than the information before, there were some exceptions where he has not being accurate or expressed the needed information in an acceptable form thow. And as I see the new information has encountered the same resistance as mine did some time ago so I appeal to Angelic Wraith again, with the suggestion that we are here to make this article better, not to preserve in its original form, edits are to be improved not removed, if you cannot face new information or it confuses you when you do face it, please leave the editing to others. And also... talk, argument edits, maybe I abuse it here since form the moment I registered this talk section got approx. 10x bigger, but I'm pretty sure its made for this.--St4r4m4m4 03:38, 15 February 2006 (UTC)

(btw all that pseudo-grammar stuff about either "8MB" or "8 MB" is starting to look funny, in a not too pleasant way, I'm no one that can make orders here, but anyway, I demand that the next one who makes an edit of that kind here MUST argument it so we all learn the "secret" about these terms, and no further (useless) edits are done regarding this)--St4r4m4m4 03:38, 15 February 2006 (UTC)

No, you're not abusing this discussion- as I understand it, this is exactly what it is for, and a discussion is a definite good thing :)

I don't know what you're referring to whne you talk about "pseudo-grammar.." the "8MB and 8 MB?" I realize it's starting to look a little funny. It should all read xMB and xMHz (as all professional sites do), but Wikipedia's policy sucks on this part, and it says that it should have a space for "readability," which is crap if you ask me. A certain person goes through pages and puts spaces between all of the numbers and specs (i.e., he changes 128MB to 128 MB, and 800MHz to 800 MHz). I'm just trying to slowly change it all back (on this page only, really) to 128MB and 800MHz. Lol. Anyway, what "terms" are you referring to? I'm confused.

Dan 04:32, 15 February 2006 (UTC)

Voodoo 3 4000

I'm pretty sure the V3 4k was supposed to eventually come out- it was certainly talked about a lot... but obviously it never did. I have a vague recollection that it may have turned into the V4 4.5k, though obviously it would have had to switch to the VSA-100 chip..

In my personal opinion the V3 4K is just and only a pre vsa100 speculation, there is no prototype or preproduction sample of such card as far as I know, the speculations found around clearly point to vsa100 capabilities (4xagp, 32bpp) Such card name is just a product of speculations, so I belive it clearly does not belong in the chip table, all other mentioned cards exist in form of prototypes and are more or less known to be "confirmed to exist" by 3dfx or other reliable 3dfx company related sources.--St4r4m4m4 15:13, 14 February 2006 (UTC)


- Response: Guru3D had a snippit about this in an article on the V3 series. The V3 4000 was either a marketing ploy or a community rumour. It was never planned. NitroX infinity 07:45, 7 June 2007 (UTC)


sorry if i jump in the discussion, however i remember that sometimes 3dfx cards, when launched on the market, had different number than what appeared before in the press.

IIRC v3 4k is just the preliminary name of v3 3500. at the beginning, they were thinking to sell the cards as voodoo3 2000/3000/4000. in a second time they changed 4000's name to 3500 (maybe because of the 3 in name, like voodoo 3 and 3dfx)

the same thing happened with vsa100 cards: at first, the names were supposed to be 4500, 5500 and 8500 for the 4-vsa100 card. this was later called 6000 (although it did never arrive to market)

all of this IIRC. and sorry for my bad english. —Preceding unsigned comment added by 82.50.117.150 (talk) 12:54, 11 October 2007 (UTC)


8500 was supposedly to be an 8-chip card. NitroX infinity 14:53, 27 October 2007 (UTC)

Voodoo 4 4500 AGP 4X support

As a moderate 3dfx collector I must say that the info about V4 4500 not supporting AGP 4x is simply wrong, AFAIK this is the only 3dfx card whose board supports and works at AGP 4X mode (excluding from the statement non-3dfx boards with single VSA-100 chipset) If there is any disagreement please doublecheck before re-editing again.

Some references: http://www.falconfly.de/identify-3dfx.htm

http://tiger.towson.edu/~tzeger1/3dfx/index.html (>products>voodoo4)

http://guru3d.com/voodoo4/voodoo4.htm

(the last two pages show the same statement, so it's very probable that these were actually some official specs.)

I see someone has not done his homework, and on top of that is also being unpolite.

Angelic Wraith m (Reverted. Use your brain.)

Please do doublecheck and name your sources (for your sake)

I don't want to offend anyone but the source Angelic Wraith gave us is somewhat contradictory and has a few wrong facts mentioned (not only the V4 AGP 4x support issue) http://www.xs4all.nl/~jfeelder/voodoousers/guidev45faq.html

I give enormous respect to the great company 3dfx was, so I wouldn't like people to be given misleading information, Voodoo 4 4500 AGP is de facto fully 4x AGP compliant both in voltage and transfer rate (1.5V and 66MHz quadruple data rate aka quad pumped)

Angelic Wraith you are welcome to comment and/or edit the chip table (since I don't want any flaming, I leave that to you)

--St4r4m4m4 00:05, 24 January 2006 (UTC)

My apologies. I'm used to reverting strange vandalism, and your name just looked like a sock pupput account to me- I see that you were right. Though I'm still confused. The site you gave me (http://guru3d.com/voodoo4/voodoo4.htm) lists both the Voodoo 4 and 5 as supporting AGP 4x. Still, I don't really understand why the V5s wouldn't have AGP 4x support while the V4s do.. not that it would have impacted performance for those cards, but it does look better on paper. Plus there's the AGP 2x/Pentium 4 issue...


Actually I noticed there is a lot of "wrong" info on the net regarding specifications, not to be misunderstood the info on eg. guru3d.com and other "wrong" sources are actually based on official pre-release specs, wich obviously didn't turn out so at release.

So let's clarify a bit on these support issues...

(people non familiar with AGP specs read the Accelerated Graphics Port article for better understanding)

The VSA-100 chipset by itself is made for supporting 4x AGP

The cards:

Voodoo 4 4500: fully 4xAGP compliant (physically, by voltage, by transfer rates) note: non 3dfx single VSA-100 cards have 3.3V signalling so don't get them burned, only the 3Dfx card has full 4xAGP

Voodoo 5 5000: 2xAGP (most of the prototypes seen are physically compatible with the 4xagp slot, but not in any other way nor voltage nor trasnfer rates, so if anyone has the luck to come up with one don't run it in 4x mode, it will get burned, not necessarily the moment you start the pc so be very careful)

Voodoo 5 5500: 2xAGP (incompatible physically, electrically and by transfer rate, afaik some prototypes might appear having physical capabilities for a 4x slot but it's the same as the 5000, dont get it burned it just won't work)

Voodoo 5 6000: 2xAGP (fully incompatible, no revision to be seen with any kind of 4x support)

About why is that so, I belive since 3Dfx was in a hurry to deliver the VSA-100 product family to the market they did not manage to implement 4xAGP on all cards, the Voodoo 4 4500 being the simpliest got the 4xAGP support on time.

All other older AGP capable 3dfx chipsets and cards are not 4xAGP compliant in any way

Rampage prototypes (those that made it to work) are fully 4xAGP compliant

Some extra on top of this about P4 as you mentioned, the main thing is that no motherboard chipset that supports P4s also has 2xAGP support so most of the agp voodoos are not usable with P4 processors. The latest chipset supporting 2xAGP is VIA KT333 supporting AMD processors.


I'm quite new to wikipedia editing (I mostly read it so far) but since I put a great deal of my time into 3Dfx related stuff I decided I will make this article better as soon as my exam period is over :)

--St4r4m4m4 12:58, 24 January 2006 (UTC)


- Response: Why V5's don't work on AGP4x? Because apparently the AGP4x signals interfere with the VSA-100's SLI signals. NitroX infinity 07:53, 7 June 2007 (UTC)

SLI edit



This is what infuriates me about Wikipedia. Angelic Wraith reverted my recent edits to the SLI paragraph by saying: "It was less confusing and wordy the other way". Well I respectively disagree. As I was reading the article the last two sentences of that section were not clear to me, so I took some precious time out of my day to try and improve them. Were the changes perfect, no way. I probably would have completly rewritten the section if I had time.

I would have no problem if the section was re-edited with new words but to simply revert the changes dismisses the fact that at least one reader had a problem with them.

The way this last sentence is worded implies that on screen output is what is different not the algorithms:

While this shares the same name, the algorithms for sharing the scene between cards and then recombining the end results are totally different.

Now I am going to revert my changes and see what happens.

= AGP 4x

Also I dont believe this claim (but I need to research this before I change it)

It was also limited to AGP 2x, not supporting AGP 4x, making it incompatable with Intel's Pentium 4 processors.

I have modern motherboards running ancient AGP cards (Riva TNT and Matrox Millenium) just fine. The AGP standard is backwards compatible AFAIK.


--bpatton 25 January 2006

Yes, well, I wrote that part in the first place, and as I see it, you made it a bit more confusing, as well as less specific- and even if you hadn't, what you did was reword it to no real advantage.

You have "modern motherboards" running "ancient AGP cards." Well, that's nice, but we're talking about the AGP 2x problem specific to the Pentium 4 in combination with the Voodoo series of cards, not nVidia- or Matrox cards.

So far all you've done is tell me that you're upset that I didn't agree with your changes to the way I worded it, you haven't told me why.


I thought it was pretty clear my problem with those sentences. And yes I thought my edits improved things somewhat or else I would not have made them. I think if you dont like the edits you should propose something even better than the two versions. Othwerise my edits should be allowed to at least be viewed and edited by other people and not just rolled back by some self-appointed owner of this page.

-brad

Also I dont believe this claim (but I need to research this before I change it)

Thanks for the complain (I know lots about 3dfx but not about P4, in fact I never had one :P) I checked all the oldest P4 chipsets (VIA P4M266 | Intel 845 | SiS645) and I came up with the info that the SiS645 also supports 3.3V AGP signalling. So I take back my statement about full non-support. Changed that myself.

The AGP standard is backwards compatible AFAIK.

But still I have to oppose this statement, read the Accelerated Graphics Port "Compatibility" section, also be warned there are cases of motherboards that phisically have universal agp slots but don't support all modes. The 3.3V/1.5V AGP issue is a serious one, people HAVE burned cards due to this. Please whenever you wish run a 3.3V 1/2x AGP card (that you care about) in a newer mobo check werther the chipset & motherboard support 3.3V AGP signalling, until you do assume that it is not compatible. st4r4m4m4 is always better safe than sorry

--St4r4m4m4 16:33, 27 January 2006 (UTC)

Ok.. maybe it wasn't uniform incompatabilty.. but why would 3dfx have made a statement that they weren't compatable if it weren't true? Dan 05:28, 30 January 2006 (UTC)

Please show me that statement, I never heard about 3dfx saying that, I just heard rumors about no P4 chipset supporting 2x AGP, that's why I just posted that info only in the talk section.--St4r4m4m4 18:12, 31 January 2006 (UTC)

Rush 2D chips

As far as I can tell there is no card with Cirrus logic 2D chip, and I can also tell there is no clear distinction about any v1s and v2s, also mem sizes and clock rates are not related to any of these versions.

Facts I can tell:

2D chips:

Alliance AT3D
Alliance AT25
Macronix MX86251


The last 2 chips are found in all 4,6 and 8MB configurations, so far I have seen AT3D boards only in 4 or 6MB configurations.

Afaik there are only 2 models with 8MB:
Hercules Stingray 128/3D P/N:S3318TV (AT25)
CardExpert(Gainward) Dragon 2000 P/N:GW707 (MX86251)
The Dragon 2000 model can be also found under the brand name Fantasy FX 2 resold by 2the MAX.

Cant't tell nothing for sure about clocks, the only thing I heard about that is that the latest (8mb) version by hercules has 10% faster clock (--edit-- ok this last part is already in the article, didn't notice :P), I managed to get my hands on one sample but it turned out to be zapped :P
--St4r4m4m4 17:15, 27 January 2006 (UTC)

My understanding is that AT3D and AT25 are the same silicon--AT3D was a rebranding used occasionally on the Rush cards ("hey, look, it's 3D now!"). The chip on my card was labelled AT3D, announced itself on boot as AT25, and showed up in lspci as "AT25/AT3D." Another interesting fact is that the 6MB cards apparently had different memory splits based on manufacturer--some were 2MB framebuffer/4MB texture, some 4MB FB/2MB texture. --Jnik 20:45, 1 May 2006 (UTC)
There is some evidence of a Voodoo Rush with a Cirrus Logic 2D accelerator. It's not exactly well known though. [1][2] --Swaaye 21:19, 1 May 2006 (UTC)
The here menioned Joytech Apollo Rush got released with MX86251 2D chipset, as far as I can tell the Cirrus Logic was just planned to be used ad the default 2D chipset, but in the end it more or less never got used. I would dare say that all documents mentioning CL are all pre-release specs. Still noone has proven the existence/inexistence of these cards.
The only Rush's that really shipped to ANY partners and made it into the channel has the Alliance parts on it. Very few Macronix shipped - the 3dfx drivers barely supported those. Any work with Cirrus Logic was NOT supported by 3dfx. It is possible a board partner did this on their own. —The preceding unsigned comment was added by 68.4.29.121 (talk) 01:56, 3 March 2007 (UTC).
3dfx did build the Voodoo3 drivers to support Macronix, but again only one partner choose to ship it. Don't forget Integraph who shipped Voodoo Rush on a single planar board first. All other Voodoo Rush designs at the time had the idiotic daughterboard for the Voodoo Rush chip + memory. That alone killed some of the clock performance to the 2D controller. 3dfx had many 2d driver issues with the Alliance chip but they had source access, in fact a group of ex-ATI engineers who were under contract to 3dfx fixed most of the Alliance 2d problems prior to mass production ship by Intergraph. The only reliable Voodoo Rush board at the time was the Intergraph board as it shipped with that bug fixed Alliance 2d driver set. The Hercules board and others that shipped all at some point rolled in that newer driver codebase. A better combination at the time would have been a single AGP board (or PCI) that would have had a 2D chip JUST for 2D and emulated 3D in a window and Voodoo 1 or Voodoo 2 on it. Basically a single board with an internal pass thru. This would have been compelling enough for customers who wanted a reliable 2D board but wanted a high performance 3D board. 3D in a window was the main reason Voodoo Rush was done as it was and it was a feature that in the gaming space is barely used today. (VRML, and other 3D apps were thought to be killer apps for 3D in a window at the time)

Voodoo 5/Voodoo5

I was always under the impression that the cards were named Voodoo3, Voodoo4, Voodoo5, without the space. There are a lot of pages which corroborate this.. but then there are those which do not. Any thoughts?

Voodoo 5 6000

Although, somehow A couple hundred voodoo5 6000's did manage to get into the market for the diehard 3dfx fans.

Fact is Voodoo 5 6000 NEVER got retailed, not in one sample, the cards of that as you mention diehard 3dfx fans are completly ALL prototypes, no exception, how do ordinary people with a 3dfx hobby get such valuable prototype is another story. If you wish to rewrite that I suggest you to replace "into the market for" with "into the hands of" (a reminder and my opinion too, this is an article about the 3dfx company, not about 3dfx fans, if you are a fan dont't misunderstand me, I'm a fan myself, but an article about something should still have the information about that something) --St4r4m4m4 17:55, 31 January 2006 (UTC)

To Angelic Wraith, about "models" or "dates": In the matter of fact those numbers (not at all a matter of grammar) are simple PCB dates (in the format WWYY) so stating "model" before them is not exact, if you want to make some corrections I would suggest a combo like "model date" or "revision PCB date" or maybe simply "PCB date" or "revision date", make your pick the choice is good :) --St4r4m4m4 00:26, 2 February 2006 (UTC)

Prototypes

I see a lot of info about prototypes appering in the article, and I'd like to suggest, to remove all of it and create a proper Prototype section in the article. As I see it at this rate the article will be filled about all kinds of prototypes in every section of the article. There is a huge load of various prototypes and/or rare special cards. (also the whole article is a mess... eg. why the heck are V3,4 and 5 subsections of the Decline section???) Don't look bad at me for complaining all over, I'm willing to make a bigger edit & cleanup but I have to take care about real-life stuff first (exams) --St4r4m4m4 18:23, 31 January 2006 (UTC)

Well, ya know... 3dfx started to go under with the V3s and seriously tank with the V4/5s..

If is that that you want to tell me, then the article is self contradictory. (and u can't just tell "ya know..." you were not in the company at the time, there are all kinds of speculations (market shares, delays, agreements, policies, technologies, managment ops decisions...) so discussing the cause like end-users would probably be unproductive) The "Cause for Decline" clearly states about this issue already in the first sentence. I fully agree with the statement of unclarity of the exact cause of the decline, as I see it it was more a series of various events that led to the end of 3Dfx, not a single single decision, eg. the changes in the V3 market strategy had nothing to do with the dreamcast project or the delay of the VSA-100 product family. Anyway you are somewhat right if you look at the timeline someone might consider it so, but my intention was more to point out the article's construction (In my personal opinion when I look at the context box I only do like "WTF?") Information seems poorly arranged trough all sections and paragraphs. (btw please use signatures in the talk section :S it's easyer to read it trough) --St4r4m4m4 00:18, 2 February 2006 (UTC)

Mac Editions

Shouldn't we add something about the Mac editions? They had some pretty neat products from now-defunct 3rd-party developers.. plus we just need to mention the Mac Editions anyway.

Good point, 3Dfx made some nice contributions to the mac community, older cards by VillageTronic, Techworks newer by 3dfx itself. (I'll consider that when doing bigger edits)--St4r4m4m4 00:33, 2 February 2006 (UTC)

I think the first official Mac Edition was the Voodoo4s and 5s... both PCI-only.. but there were cards like the MacTell Evileye 3500 (Voodoo3) 183/183MHz... there was even a card for the original iMac to fit in its proprietary Mezzannine slot... some Voodoo2-based card... made by GameWizard I think? 8-12MB of RAM... Dan 04:46, 2 February 2006 (UTC)

VillageTronic did a bulk of the real engineering effort to get Mac to work properly. Techworks was nothing more than taking existing SW and packaging it up.

Rampage (Spectre)

We should talk about the Rampage and its obvious implications.. hardware T&L... 2x200MHz minimum, 2x250MHz without much trouble, VSA-200, DX8.0, 0.18µ process... etc. You know, there were not only articles on it, there was a site which HAD one which was going to benchmark them.. I can no longer find any references...

Speaking of Rampage and further 3dfx chipsets in which I should include the Sage geometry engine, both belong to the Spectre product family, having the following combinations in cards: Spectre 1000 one Rampage chipset, S2000 one Rampage and one sage chipset, S3000 2xRampage 1xSage, and the S4000 4xRampage 2x Sage which is AFAIK only speculation. All the above mentioned is mostly theorizing, except of course the Rampage chipset, as you mentioned 200-250MHz with an estimated fillrate 800 to 1000MPixels/s, 4 pipes with 1 TMU per pipe, 2-4X FSAA. The Rampage chipset made it to live a short period before 3dfx closed its doors, 2 kinds of bringup boards were known to exist, each variant in 10 samples, most of the samples were not working, some got in the hands of 3Dfx collectors as you mentioned. Benchmarking a Rampage is likely to show not too great results since the lack of good drivers (or should I say the lack of any kind of non alpha/beta stage drivers) the referred benchmarking is more like an effort to make it work at all, and run some popular 3D applications on it. --St4r4m4m4 15:14, 7 February 2006 (UTC)

Yeah... I dunno. I think the Rampage had a flipped DAC, and needed a dongle to make it work properly. I have a lot of trouble understanding what you're trying to say.. could you start punctuating your sentences? I don't mean to be rude, but it makes it pretty difficult to make sense of.

Btw, I saw what happened to the chip table. What the hell. That kid made it look positively terrible, ruining the page, and he did it in like thirty edits. WTF? Dan 01:30, 15 February 2006 (UTC)

Added some puntuation, If it does not please you, its the best I can do at 4am. My punctuation might be quite bad and my english not flawless, but still I don't use WTF-like acronyms for commenting about edits, and this "kid" was probably trying to improve the information in the article not spoiling it, and you cannot say you expect that all wikipedia users are higly experienced with editing an article or have some taste about table looks. If you think you are better handling form and style please do that, for more read my comment on top of the talk section.--St4r4m4m4 04:01, 15 February 2006 (UTC)


You seem to misunderstand me. I'm not making fun of you, or putting you down. I was having actual trouble understanding your meaning, and there's really no other way to say it, is there? Btw, WTF isn't technically an acronym, it's an initialism :D (sorry, couldn't help it). ;-)

No, I don't expect all wikipedians to be "experienced..." in fact, I don't really even know how to make a table, I just know how to edit one, since all I have to do is analyze it. As I said, I really appreciated your turning all those asterisks (*)s into superscript numbers, it looks much better that way- so I'm not totally opposed to change ;). You seem to take everything I say as a personal attack on you, and I'm sorry if it comes across that way, but I don't mean it to. I personally greatly appreciate your effort you put into the page, and the knowledge you have of 3dfx that you add. I didn't appreciate that person's edits, and I personally can't stand it when people make lots of little tiny edits in a row.. it drives me crazy. Hope this clears things up a little. Dan 04:32, 15 February 2006 (UTC)



- Response: 1) It's Specter, not Spectre. 2) What's the original source of this Specter 4000 card? I've never seen it been mentioned before.

Also, quit the VSA-200 nonsense, that chip is a myth. Rampage would have never been called VSA-200 because it's a completely different architecture. And Daytona is a VSA-101 chip.

NitroX infinity 08:08, 7 June 2007 (UTC)

Velocity series

Seems nobody here ever heard about such thing. Show me the best you can do, I'm too busy now.--St4r4m4m4 15:16, 14 February 2006 (UTC)

The velocity series was marketed by STB and were just stripped down versions of the Voodoo3. If I recall correctly, they had one less texture unit and half the memory of a stock Voodoo3. I think the V128 ran at 143 MHz, though I'm unsure as I have not powered my V128 up in years. 24.206.116.9 19:28, 24 August 2007 (UTC)

Velocity's have 1 TMU disabled by software. With a simple registry editing the second TMU can be enabled. Both Velocity's run at 143MHz. The V100 has 8MB SGRAM and the V200 has 16MB. NitroX infinity 10:54, 11 September 2007 (UTC)

omg

I corrected a lot of things at night, but u placed back the bad articles (wrong specifications, never existed cards, ect.) u sh*tted back the whole article. The chip table looked good, i dunno what was your problem with it, but i say if u never seed a f*cking voodoo then dont write stupidity to a global article. -Geri

I don't really have a clue what you're trying to tell me, to be honest, other than you thought the table looked good. Maybe you're using a different browser, and it looked all right on it. I'm using Firefox at 1200x800 resolution... what's your setup? Maybe it looked better on Internet Explorer, or something.

My main issue was that you seem to have added so much that the table was lopsided, uncentered and not uniform, adding extra information in some places, but not in others, and the overall effect was that it did not add to the page. Also, you added some unclear terms, like "SG," as to which I had no idea what you meant. I'm also frustrated when people do many edits in a row, since it confuses things and is hard to follow, as well as takes up space and makes things harder to track over time. Maybe we can come to an agreement on the discussion page, and then we can add to the table. Dan 21:41, 15 February 2006 (UTC)


Hi! I use firefox at 1024x768 res! SGRAM, SDRAM and EDO RAM are RAM types. -Geri

Oh, SGRAM. Yes, well, sort of. SGRAM is basically slightly customized SDRAM- more expensive, and (obviously) used in video cards. Just saying (SG) doesn't make a whole lot of sense; I've known what SGRAM is for years, and had on idea what you meant. Dan 00:45, 17 February 2006 (UTC)

In my opinion it's ok, stating "SG" along a mem quantity of a video card makes it quite obvious to people who know about hardware.--St4r4m4m4 18:55, 17 February 2006 (UTC)

Lol... no it doesn't. SGRAM does... SG doesn't. Either way, it's mostly irrelivant whether or not it used SGRAM. P.S., this article is for ANYONE, not for people "who know about hardware." If we're talking about those types of people, why even have the page at all, if they know everything anyway? Like I said, I've known about SGRAM for what... eight years? And I still didn't make the connection. Dan 04:07, 20 February 2006 (UTC)


Voodoo Banshee 110MHz RAM..?

Go here, and click Products-->Voodoo Banshee. http://digilander.libero.it/F1Land/3dfxarchive/ It says it was 10ns EDO RAM, running at 100MHz (10ns RAM is only capable of 100MHz, right?)


True 10ns RAM is not capable (unless in wierd circustances) to run @ higher clock. Banshee part of the answer... cards with this chipset exist in more variants, true some have 100/100 clocks and 10ns ram and are not uncommon, But still the majority of Banshee cards are equipped with 8ns RAM and run @ 100/110 clocks, with this in mind I also confirm that Banshee is the only 3Dfx chipset with different core/mem freq. No sources to show, but I can tell you I have about 20 different Banshee cards in my posession :S --St4r4m4m4 20:58, 24 February 2006 (UTC)

8ns is spec'd for up to 125MHz operation.. obviously.. since 8ns/1s= 125m cycles per second... and the site says 100-125MHz memory operation... >_< Dan 21:45, 24 February 2006 (UTC)

I'm not positive anymore, but I seem to recall that some boards back in those days did run RAM out of spec. It usually worked fine because they tested the boards at the speed. But I recall there may have been some problems too with such boards. Usually overclocked and misbehaving video RAM gives you specks of wrong-colored pixels on your desktop.
Banshee used SGRAM I believe, not EDO. EDO ended after Voodoo2.--Swaaye 21:15, 1 May 2006 (UTC)

Three-dee eff ex

Is it 3DFX or 3Dfx or 3dfx? This and several related articles use all three interchangeably. The early on-chip logo clearly says 3Dfx, whereas the later logo is 3dfx. I assume the answer is 3dfx. -Ashley Pomeroy 01:47, 8 January 2007 (UTC)

The original logo was 3Dfx, and recieved a redesign with the release of the Voodoo3 to 3dfx Draknfyre 20:01, 15 February 2007 (UTC)

It was 3Dfx all the way up until the company rebranding around the time of the STB acquisition (or disaster). At that time marketing made it 3dfx In any case press, customers, etc always had their own style.

Glide API

I cleaned up the section on Glide a bunch. Glide was designed around the hardware, not the other way around. And Glide has little to do with OpenGL. A full OpenGL implementation was *not* available for the Voodoo Graphics or Voodoo2, that's why the "MiniGL" driver existed. As currently placed and written, this sections sounds like its primarily pertaining to the original Voodoo Graphics, and I've retained this for now. I no longer remember if the Voodoo3 or newer ever got a full working OpenGL, I had traded in my Voodoo Banshee for a TNT2 Ultra by then... :)

Its still greatly oversimplified and basically wrong, but it irks me much less. Nearly every manufacturer of the time had a specialized low level API for their chips, the developer preference towards Glide had more to do with the superiority and thus popularity of the Voodoo hardware itself than anything else. OpenGL was designed for workstation hardware, which had been capable of such things as hardware T&L for some time already, and thus was overcomplex and overkill for consumer hardware at the time, not to mention the difficulty and expense of first writing a full OpenGL implementation, *then* getting it approved and licenced by the ARB... As consumer hardware began to implement features such as T&L, OpenGL became more relevant. And Direct3D, well I'm not even going to go there... 66.216.226.167 07:43, 18 March 2007 (UTC)

Scandal

Nvidia found some stolen intellectual property in 3dfx products, and the lawsuit settlement ended in selling out. This is how the news reported it when it happened. I don't see anything in the article about this.

- Response: To my knowledge, 3dfx sued nVidia over this. There was no lawsuit settlement. nVidia figured; why pay a fine of $4 billion if the company (3dfx) is only worth $400 million? That's when nVidia bought out 3dfx. NitroX infinity 08:12, 7 June 2007 (UTC)

Memory

EDO-DRAM (Voodoo Graphics, Rush and 2)
SDR-SDRAM (Voodoo Banshee, 3, 4 and 5)
SDR-SGRAM (Voodoo Banshee and 3)
DDR-SDRAM (Daytona (VSA-101))
DDR-SGRAM (I believe the 16MB Daytona cards have this.)

!Voodoo 4 4000 is speculated to be a VSA-101 chip (Daytona) with SDR-SDRAM
!Voodoo 5 5000 prototypes also have SDR-SGRAM

NitroX infinity 08:20, 7 June 2007 (UTC)

Reformat

There is a lot we know about 3dfx. Too much to put into one article. Perhaps rebuilding all 3dfx related pages into the following format? :

Pagename: 3dfx

Introduction of cards, acquisition of other companies (STB, Gigapixel, etc.), legal actions (the nVidia thing), financial deals (SEGA for instance) Would contain a timeline like on Gary's site.

Pagename: 3dfx Hardware

About the cards available in retail. All those loose articles (v3, v4/5, etc. would be merged into this one).

Pagename: 3dfx Prototypes

About the prototype hardware

Pagename: 3dfx Merchandise

There was more than just videochips. 3dfx had a lot of merchandise.

NitroX infinity 08:39, 11 June 2007 (UTC)

Actually I think the content is fine for the most part. Reading over it now, my only concern is the "failure" area, which is repeated twice (in effect) and seems overly long. Perhaps a little cleanup applied there would address the length issue too. Maury 12:20, 11 June 2007 (UTC)

3dfx was not acquired

From Nvidia's website: "NVIDIA acquired certain assets, technologies, patents and intellectual property from 3dfx, however, we did not merge with 3dfx nor did we purchase the company. All liabilities, including drivers and customer support for 3dfx Voodoo, Voodoo TV and STB products, remain the responsibility of 3dfx." Nvidia's website

So who owns 3dfx? Did it dissolve (can corporations dissolve?) Is it technically still a corporation but one that just doesn't do anything?

Af1218 23:24, 4 July 2007 (UTC)

This is rather interesting... It was apparently last edited on May 9th, 2007. Corporations, when they are going under, do often retain skeleton elements to deal with the rest of the legal banter of sales and such. I believe that 3dfx no longer operates as a true company, but merely as a skeleton element in negotiating the sale of its assets to other companies. This means that negotiations over the sales of 3dfx's assets to nVidia still have not resolved. - XX55XX 18:49, 8 July 2007 (UTC)
Hmm, I think something very interesting, or perhaps illegal (?) has been happening recently with 3dfx, and I figured I'd come here to see if anyone knew. It sounds like the discussion above might be part of the answer, as I believed up to now that 3dfx was completely defunct. However, I haven't seen them do *anything* in... what, almsot a decade? Anyways, I noticed yesterday that 3Dfx's stock still exists, and has gone up 33% since July 3rd (though obviously in that scary "penny stock" zone where much mischief occurs). Is 3dfx alive as a company still/about to announce some kind of new business venture, or is someone playing games with the stock? If there's new company activity, what is it? If there's not, should someone report this to the SEC? - Anonymous 20:19, 9 July 2007 (PST)
Annoymous, do you have any links to 3dfx's SEC filings page or to its recent stock queries? - XX55XX 18:57, 13 July 2007 (UTC)
Nay, I noticed it when I was tracking them on Google Finance. The symbol's TDFXQ. Enter it in and you'll see what I mean. *someone* is buying/selling it, but I don't see any information on activity other than some old finance sheets (http://finance.google.com/finance?fstype=ii&q=OTC:TDFXQ). I personally think some funny business is going on. There's no reason someone should be buying and selling stock in a defunct company. It also looks like they've gone back down a few cents the last couple days, but the whole thing smells bad to me. - Anonymous 19:29, 18 July 2007 (PST)
(Anonymous2 speaking) Well, taking a look at the SEC filings seems to indicate a number of different funds have owned 3dfx since then. However, none have reported any profits which is not surprising now that they're headquarters is a P.O. box which can be found in Item 1b of the 2008-2-14 filing. They still have lots of liabilities which have never been paid off. It appears that 3dfx cannot be truly considered dead until all stockholders decide to dissolve the firm which never happened. My most likely suspicion is the current investors are hoping for another company to want to buy the name at which the buying company would settle the remaining liabilities plus extra for profit for current shareholders and then transfer ownership to the buyer. Of course, I have no proof this. But the only thing 3dfx is at the moment is a name and some liabilities. 149.169.122.130 (talk) 00:34, 22 September 2008 (UTC)

Form 10-K for Fiscal Year End January 31, 2002 and sec.gov filings, the latest being from 2007-05-17 when I checked. Hope it helps. Ǣ0ƞS 00:38, 16 August 2007 (UTC)

Thanks for that link. It perhaps seems that the people owning 3dfx are part of some venture capital group. Wonder what they are trying to pull off here? - XX55XX 20:38, 24 August 2007 (UTC)
It seems that 3dfx was in the process of suing nVidia as of December 9th, 2005. If this litigations is unresolved then the potential for profit or loss resulting from it could be part of the reason that 3dfx is seeing its stock change hands. See page 15 of this December 9th 2005 filing for hints of ongoing litigation and see this page or this one for what I think is all SEC information on nVidia. Af1218 02:57, 27 September 2007 (UTC)


The stockholders own 3dfx but the trustee appointed by the US Bankruptcy Court controls it. The trustee sued Nvidia and lost. See reference 2, Court's April 30, 2008 decision. The Official Committee of Equity Secuity Holders also sued Nvidia, requesting a judgment that 3dfx is owed six million Nvidia shares. That suit, Adversary Proceeding 06-05115, has a status conference on August 28, 2008 in San Jose.

The six million shares started as one million in the December 15, 2000 agreement between 3dfx and Nvidia before splits. If the Nvidia shares are $10 each, then after fifty million dollars of bankruptcy expenses, each of the forty million 3dfx shares is worth 25 cents. Victoryjc (talk) 10:24, 24 July 2008 (UTC)

The Voodoo Lists

Information is good, so perhaps a link to my lists could be included in this article?

http://www.falconfly-central.de/cgi-bin/yabb/YaBB.pl?board=general;action=display;num=1165598928

My advice, should these be included, is to use the above link. It not only has links to the newest files but also tells when updates were made.

NitroX infinity 20:51, 9 September 2007 (UTC)

Why were all the links removed?

Why did Gimme Danger remove all the links to external sites? NitroX infinity 11:16, 5 October 2007 (UTC)

I'm working with WikiProject External links to ensure that the external links in articles comply with WP:External links. In particular, I'm removing commercial sites, forums and suggesting that most external links be incorporated as references into the article. The links I removed from this article fit these criteria. If you feel that a link fits external link requirements, feel free to add it back. --Gimme danger 11:58, 5 October 2007 (UTC)
Ah, I understand now. So basicly, only links to unbiased information should be present? NitroX infinity 09:34, 8 October 2007 (UTC)
Yep. More detailed information is available at WP:EL. --Gimme danger 17:56, 8 October 2007 (UTC)
May I suggest the following links then:

http://www.falconfly.de/identify-3dfx.htm (to identify hardware by partnumbers)
http://welcome.to/3dfxBios (3dfx BIOS files)
NitroX infinity 11:28, 10 October 2007 (UTC)

'Only 3D' looks like a typo for '2D' ?

"2D/3D - products released before the Banshee can only display 3D graphics.. " I'm not bold eough to fix it ! --195.137.93.171 (talk) 10:34, 10 December 2007 (UTC) No, this is right. The cards only worked in fullscreen 3d modes, they couldn't render things like the windows desktop.j

This is incorrect as 3Dfx Rush cards were pre-Banshee and were 2D/3D cards back in 1996.

Adderek (talk) 21:42, 3 May 2010 (UTC) Every product that can display 3D is capable of displaying 2D. There was a video player for Voodoo2 using Glide. However we might consider a modification of the description. Graphic cards based on Voodoo (1) and Voodoo 2 chipsets were designed for 3D acceleration only while 2D display should be provided by a separate dedicated graphic card.

Image copyright problem with Image:Unreal-GlideVoodoo1flyby.jpg

The image Image:Unreal-GlideVoodoo1flyby.jpg is used in this article under a claim of fair use, but it does not have an adequate explanation for why it meets the requirements for such images when used here. In particular, for each page the image is used on, it must have an explanation linking to that page which explains why it needs to be used on that page. Please check

  • That there is a non-free use rationale on the image's description page for the use in this article.
  • That this article is linked to from the image description page.

This is an automated notice by FairuseBot. For assistance on the image use policy, see Wikipedia:Media copyright questions. --02:48, 4 November 2008 (UTC)

nVidia resurrecting 3DFX brand?

On January 12, I reached a site which mentioned the new GeForce 3D Vision products by nVidia (which I am aware they exist), but it also mentioned some things that seemed "suspect" to me. Unfortunately, I lost the link to that site, so I cannot show it to you directly.

Anyway, one of the things it showed was a video card with an orange (and not green) nVidia logo, calling it "a video card specifically made for stereoscopic 3D". But that's not all. It also showed a peculiar logo on the GPU of that video card, stating that nVidia introduced "a totally new brand" to distinguish products made for stereoscopic 3D from their "regular" products.

I made the mistake of not saving the picture which showed the logo, so, in order to show it to you, I had to draw it from scratch, trying to get it as close as I could remember. And this is a crude representation of the logo for that supposed "totally new brand".

So, did anyone read anything like that? Is nVidia really going to resurrect the 3DFX brand and use it for stereoscopic 3D-related products? Or was the only intention of whoever wrote that article to dupe clueless n00bs who never heard of 3DFX? Or was he a clueless n00b himself? Devil Master (talk) 19:25, 13 January 2009 (UTC)

  • Well, I'll reply to myself here. I asked the question at the MTBS3D forum and I've been replied that it's completely false. Moreover, looking for "3DFX" on USPTO shows no recent activity from nVidia, so I'd say that whoever posted the story and the doctored photographs on that site just wanted to play a prank. Devil Master (talk) 21:04, 13 January 2009 (UTC)

Removing refrences to time travel

The Omega Threat? —Preceding unsigned comment added by 24.225.22.231 (talk) 13:43, 5 March 2009 (UTC)

intro paragraph: "one of the most high-profile demises"?

"one of the most high-profile demises in the history of the PC industry?" Honestly, by what critiera? At least several other PC-related companies either went bankrupt (Packard Bell), or were acquired (ATI), that had higher reveune/marketshare at the time of their business transition. In terms of PC graphics suppliers, ATI is the largest to have experienced a change of ownership. —Preceding unsigned comment added by 99.182.67.220 (talk) 15:00, 6 June 2009 (UTC)

There is a difference between an acquisition and a demise. In the case of ATI, AMD purchased the entire company as a unit, and not as a result of ATI's failure as a company, but rather due to ATI's success in the discrete graphics segment. In 3Dfx's case, the company outright failed on a financial level, and only certain assets were purchased by Nvidia as part of a bankruptcy creditor settlement. Nvidia didn't acquire 3Dfx "the company", they acquired 3Dfx's technology, IP, and some R&D staff; the rest of the company evaporated. (e.g., 3Dfx's Letter to the community regarding their failure)
"Under this agreement, nVidia will acquire certain, specific assets from 3dfx including technology, company and product brands and other assets. In addition, the 3dfx board of directors has recommended to our shareholders the dissolution of the company over the next few months."
Your comparison between ATI & 3Dfx doesn't make any sense, they were very different situations. Toastysoul (talk) 17:28, 5 September 2009 (UTC)

Should it say "Rampage"?

Originally scheduled for demonstration at the 1998 Comdex event, the first functional samples of rage only debuted in 3dfx's labs in December 2000.

Should it say Rampage instead of rage? 186.136.87.139 (talk) 04:31, 31 July 2009 (UTC)

Yes, it was called rampage, not rage. [3] [4] Toastysoul (talk) 17:30, 5 September 2009 (UTC)

online video hardware acceleration

is there any way to use facilities of hardware accelerators - to watch flash etc video faster without gaps etc??? —Preceding unsigned comment added by 95.133.226.18 (talk) 21:18, 21 September 2010 (UTC)

Behind the curve

Maybe worth while to look at things 3dfx did that even though they helped short term revenue hurt the company in the long run.

Some things I can thing of:

- No real 32bit support in 3D until after Voodoo 3

- Slow to get confomrant ICD OpenGL drivers in Windows (relied on MiniGL too long)

- Glide always took precedence over Direct3D - D3D looked like a wicked step child

- Video quality in 2D/3D combo devices was sub par (DVD playback etc)

Adderek (talk) 21:32, 3 May 2010 (UTC) 16-bit picture of 3dfx was better than 16-bit picture of nVidia. During gameplay you would not see the difference between 3dfx 16-bit and nVidia 32-bit picture. The issue with 32-bit support was started by nVidia with a demo that worked only with nVidia products - so the goal was to show that 16-bits is not enough instead of fair comparison. Additionally there was "22 bits trick" mentioned at Voodoo3 that improved the display quality.

I remember quite distinctly when I blew $200 replacing my Voodoo Banshee with a TNT2 Ultra, the difference between the two was absolutely night-and-day. Very visible in the reflective floors in Unreal, for example. I have an old screenshot somewhere if you want proof. Though really, any screenshot from an early Voodoo stands out like a sore thumb to me, due to their distinct grainy dithering. 3dfx may have improved things in later hardware, but the TNT blew the Voodoo2/Banshee out of the water as far as 16-bit dithering went. I wouldn't know, after the Banshee I went Nvidia and never looked back. :) 24.179.151.124 (talk) 20:36, 22 November 2010 (UTC)
Hmm. I had a TNT2 (not Ultra) in my PC until years after it was obsolete and it never provided me with anything but headaches and disappointment. The bugs, artifacts and crashes notwithstanding, 32-bit color was useless in games that actually supported it, because the card lacked the fill-rate capability to actually run them above 640x480x32.
Also, if you're talking about the original Unreal, then you're obviously just making things up, as the original Unreal was optimized for Glide and always runs better on 3dfx cards. You're obviously a fanboy anyway, so your opinion matters little.
Wannabe rockstar (talk) 15:29, 28 October 2012 (UTC)

Note that many of the LCD displays can display only 18-bit colors so 3dfx would provide more than enough with 22-bits. Such display depth of 18-bit color results in 262144 colors available and it is described at Color depth thus the nVidia's demo should present only a small difference when used on most LCD screens.

Never mind that back around 98-99 when the 16-bit dithering issue was relevant NO ONE had LCD monitors, and laptops didn't have 3d chips in them... 24.179.151.124 (talk) 20:36, 22 November 2010 (UTC)

As a result it should not be a "short term revenue that hurt the company in the long run" but rather an "optimal solution with a small deal-off that was successfully overused by the competition for an unfair product comparison".