The Good Old Days of 3Dfx

Fosters writes: "There's a short story about the old days of the 3D graphics world, when 3dfx (3Dfx) were kings of 3D and how things have changed in today's industry. The authors talk about how that came about, albeit somewhat light-heartedly. This sums it up as the author says, "To this day, I truly believe that this was the turning point for 3dfx and their SuperG downhill slide (that's a winter Olympics event). And it wasn't because of some fancy technology, a military leader (depends on how you look at some of the former VPs at 3dfx), or even a drill instructor named Zim (yeah, Starship Troopers- too easy). 3dfx started to lose their fan-boys and early technology adopters to NVIDIA then, who were waiting and watching, as 'Bugs' do, with TNT, TNT2 and something more than '22-bits' of color.'"

Comments

3Dfx vs nVidia vs The World

[RISCy Business]

Well, they're both making chipsets. Wow.

3Dfx makes cards. nVidia does not - they simply make chips and sell them to manufacturers to make the card.

They both claim to be better than the other. Now, gee, am I the only one seeing atypical competition there?

3Dfx wants to innovate - ie; use a stagnant chipset for a base and make a better design. nVidia wants to go faster - ie; redesign every 6 to 12 months to milk out another 3FPS on systems that aren't already running into bus limitations.

Reguardless, I won't buy either. Why? Because they both claim OpenGL support. Now, I don't know about you folks, but seeing as I work in AutoCAD frequently, that means hardware support. Neither of them have it. ATI doesn't. #9 didn't. Why? Most of them view software as the future.

The mistake both nVidia and 3Dfx are making is that they're trying to take on the world when they don't have the staff, technology, or knowhow to do it. I've seen the Voodoo3, the Voodoo5, the TNT2, the GeForce MX, etcetera. The 2D quality, quite honestly, SUCKS. I've seen better 2D rendering in a blender. And I'm not talking a video blender, I'm talking the one in my kitchen. The refresh rates are for the most part inadequate for professional graphics work, the 2D image quality is abhorable at best, and considering I'm hitting bus limitations, I don't see how the extra 30FPS it's capable of in the chipset is going to help any.

I remember back about a year or two ago, I have a 3DLabs Permedia2 8M PCI card and the Voodoo had just come out. My friend snagged one, I scoffed. I said 'ha! My card already does all that stuff.' His 8M Voodoo continuously and routinely got smoked by my Permedia2 when it came to games. QuakeGL, Final Fantasy 7, etcetera. My Voodoo2's (2x 12M - for Unreal Tournament, there is no other option) combined with my #9 RevolutionIV 32M still are VERY hard pressed to beat the Permedia2 in any number of tests.

Now if you go read AnandTech sometime, you'll note that a lot of the cards these days - at least the gaming cards - are getting *OBSCENE* FPS rates. >70FPS. And when you pair the big names - nVidia, 3Dfx, Matrox, ATI, etc - up in, say, a pIII 733 or whatever it is, they ALL get the SAME RATE at 640x480x whatever depth. Why? The bus is full. Can't push any more data than that. And what are these companies doing about it?

3Dfx is adding an external powersupply for all the active cooling you're going to need just to run the Voodoo6 at normal speeds. nVidia is gleefully ignoring it and boasting a faster and bigger chip. ATI's touting more and more memory. Now, bear in mind, if the chipset doesn't use the memory for ZBuffering (mind you, not true ZBuffering), 64M of DDR SDRAM is doing you no good - 1800x1440 only needs around 14M or 16M IIRC. The companies are putting memory on the cards to make them look bigger, perhaps perform a bit better, and ignoring the core problems.

Quite frankly, I could care less whether or not 3Dfx is 'stagnating' or nVidia is 'amazing' or what have you. I need a card that works with and around bus limitations, that can do 2D and hardware OpenGL, that can do what I need. I don't buy mass marketed cards because unlike the Permedia2, which *was* mass marketed (Diamond FireGL 1000 Pro (PCI and AGP)) and an excellent card, today's cards are the equivalent of junk for me.

What do I use? Well, now instead of putting multiple cards in a single system, I'm stuck using top end cards. We're talking cards that cost more than your typical PC and more than a well configured laptop in some cases.

I just purchased, much against my desires but in tune with my *needs*, a $4,200 Wildcat 4210 graphics adapter. What is it? Dual pipeline. Dual head and a few more outputs. 90 Hz at 1824x1368. One AGP Pro 110 and two PCI connectors. All on a single card. That requires 110W of power. Wildcat was just bought by 3DLabs, the name in 'affordable' cards. (The Permedia3 is affordable, but not enough for what I do.) I was forced into spending more for a single video card than I spent on the entire system. ($2,935 for the curious. I reused the 18G SCSI-UW disks and controller.)

Now maybe some of you don't have this problem. Actually, I'm betting most of you don't. But for those of us who actually really *don't* do this for a living, per se, but need the hardware anyways (I use AutoCAD for various engine modification work on a very regular basis) are getting screwed by the dick wars between 3Dfx and company. It used to be that I could do just fine with a happy Permedia2 and AutoCAD R14. Then it was a #9 Revolution IV 32M. I went to go buy something with excellent 2D quality that could perform better than the #9 Revolution IV and found out that nothing does. If I want 2D, I have to go Matrox, which doesn't perform terribly well under AutoCAD R14 or AC2K. If I want real rendering performance, I have to go up to the professional cards, which I really didn't want to do. Now maybe the 4210 was overkill, but quite frankly, any of the cards is a pain to find and order. I could have probably gotten a 3DLabs Oxygen GVX420, but they also made the mistake of ignoring bus issues, and boom. The card ends up limited by the bus, performing really not all that much better than the other options. Just with ZBuffering and a $2300 pricetag. A single AGP/PCI combination (yes, two connectors, two PCBs) still runs into bus limits before the card hits its.

I don't know about you, but I really feel cheated.

Maybe I'll just put the Wildcat on eBay. Bidding starts at $1.

=RISCy Business

Re:3Dfx vs nVidia vs The World

[John Carmack]

>Reguardless, I won't buy either. Why? Because
>they both claim OpenGL support. Now, I don't
>know about you folks, but seeing as I work in
>AutoCAD frequently, that means hardware support.
>Neither of them have it. ATI doesn't. #9 didn't.
>Why? Most of them view software as the future.

All of the modern cards have full rasterization support for OpenGL, but I guess you are refering to geometry acceleration.

The situation has changed since you last looked at it.

The Nvidia GeForce cards have an extremely capable geometry accelerator, and they have the ability to fetch display lists either over AGP with a large bandwidth savings due to vertex reuse, or store the display lists completely in local memory to remove all vertex traffic from the bus.

The issue with professional OpenGL support has mostly been the focus of the driver writers, not the hardware. I think that Nvidia's partnering with ELSA to work on professional app certification with the Nvidia hardware was an extremely good move.

There are a few edges that the expensive professional boards still have over the nvidia consumer cards, but not many:

You can get more total memory, like a 32mb framebuffer and 64mb texture memory configuration. We will probably see workstation graphics systems with up to a gig of memory within a year. Consumer cards will offer 128mb next year, but the workstation cards can easily maintain an advantage there.

This has a cost, though: large, expandable memory subsystems can't be clocked as high as the single-option, short trace layouts that nvidia does. Even dual pipe workstation boards can't match the memory bandwidth of a GeForce2.

You generally get better high end DACs and shielding on workstation boards. The upper end of the consumer boards will do the high numbers, but it just isn't as clean of a signal.

Dual monitor has been supported much better on the workstation boards. This is starting to become a feature on consumer boards, which is welcome.

The consumer cards are still skimping on itterator precision bits. Under demanding conditions, like very large magnitude texcoord values stretched a small increment across a large number of pixels, you can see many consumer cards start getting fuzzy texel edges while the workstation cards still look rock solid.

Probably the most noticable case is in edge rasterization, where some workstation cards are so good that you don't usually notice T-Junction cracks in your data, while the consumer cards have them stand out all over the place.

Next years consumer cards should fix that.

When the consumer cards first started posting fill rate numbers higher than the professional boards, it was mostly a lie. They got impressive numbers at 640x480 in 16 bit color, without blending, depth buffering, and filtering, but if you turned on 32 bit, depth, blend, trilinear, and ran at high res, they could fall to 1/4 or less of the quoted value.

Today, there isn't a single combination of rendering attributes that will let a wildcat out-rasterize a GeForce2.

Wildcat was supposed to offer huge 8 and 16 way scalability that would offset that, but it doesn't look like it is coming any time soon.

The workstation vendors do stupid driver tricks to make CDRS go faster, while consumer vendors do stupid driver tricks to make Q3 go faster.

We bought three generations of intergraph/intense3D products, but the last generation (initial wildcat) was a mistake. We use nvidia boards for both professional work and gaming now. I still think the professional boards are a bit more stable, but they fell behind in other features, especially fill rate. Being an order of magnitude cheaper doesn't directly factor into our decisions, but it would for most people.

John Carmack

The irony that is 3dfx

[Phokus]

"The King is dead, long live the king!"

Ok well at least the first part of that cliche is true. The turning point of 3dfx is quite ironic though. After Nvidia came out with the Tnt (which was the first consumer chip to have 32 bit color), and both Nvidia and 3dfx began revealing their plans for their next generation chips, we were all surprised by 3dfx's arrogance (or naiveness) when they announced they would not include 32 bit color.

"Speed is King"

That was 3dfx's response when everyone questioned their next product, the voodoo 3. Alright so both products were relatively compareable in speed, and 3dfx still had some clout with glide games (tribes anyone?). Lessons (not) Learned?

Now what has happened? Nvidia introduces Hardware T&L on their chips, but the controversy is, who needs hardware T&L when there are no games supporting it (back then that is)?? 3dfx yet again sat on their laurels and decided to let Nvidia introduce it in their products first. While it's quite true that HW T&L was not really important back then, Nvidia was smart and marketed it as the next revolution in 3d acceleration. I mean, who wouldn't want to remove the CPU bottleneck and let the graphics card handle most of the 3d rendering? Sadly, 3dfx was only able to say "see? We have 32 bit now!"

The irony that is 3dfx

Now what do we have with the voodoo 5? Still, 3dfx REFUSES to incorporate onboard T&L when it's becomming more and more apparent that it is important these days. However, now 3dfx is ditching their "speed is king" philosophy and is trying to be innovative with their anti-aliasing and T-Buffer technology. But it seems that Nvidia has learned from 3dfx's mistakes and have included anti-aliasing technology of their own.

The road ahead?

After all these mistakes, has 3dfx learned their lesson yet? Who knows, perhaps 3dfx was right all along about not needing HW T&L right now (still, not many popular games support it) and they may very well outdo Nvidia with the release of the Napalm. But you have to admit, Nvidia played the marketing card extremely well with their Geforce cards, even if T&L wasn't really useful at the time. I pray to God that 3dfx will get their feet back on the ground, otherwise we may see another monopoly in the computing industry.

Re:nope

[John Carmack]

>Q. After reading the voodooextreme interview, it sounds like you are pursuing an allmost completely different rendering pass/phases with Doom 3. Can you give us any more details? :-)

It adds up to lots and lots of passes. I am expecting the total overdraw to average around 30x when you have all features enabled.

>Q. Could you give us your thoughts on T&L? Why does 3Dfx say it's not important?

Contrary to some comments here, 3dfx didn't just "decide not to put in T&L", the didn't have that option. Product development cycles can take years, and you can't just change your mind at the end.

They don't have it, so naturally they downplay the importance of it.

John Carmack

Re:nope

[John Carmack]

Actually, even the original Verite V1000 could do 32 bit color rendering.

At a whopping 6 mpix or so...

Rendition did a lot of things right, even on their very first card. They had all the blend modes and texture environments from the very beginning. The only thing they didn't have was per-pixel mip-mapping.

If they had delivered the V2xx series on time, they could have had a strong foothold before voodoo2. The V3xx seried would have been a solid TNT competitor, but again, it wasn't ready on time. They wound up ditching that entire generation.

John Carmack

Self-Inflicted Wound

[bonzoesc]

3dfx didn't just get mercilessly ripped apart by nVidia - they let it happen. When the TNT2, supposedly the first consumer chip faster than Voodoo2 SLI, came out, sporting 32-bit color and support for over 32 megs of video ram, 3dfx countered months later with the Voodoo3, which had the 16-bit color and 16-megs of ram limitation of the Voodoo2. That was really the first nail in the coffin of 3dfx. Their relase schedule can also be blamed, because nVidia has a new chip out every year or so, and sells them to many different board manufacturers, causing competition. 3dfx makes their own boards, just like ATI.

How is it that, in January, I bought a dang-fast TNT2 for $60, while the Voodoo2, a slower card, sold for over $100 everywhere I looked? Simple - the different board manufacturers compete with each other, trying to sell their TNT2 board over somebody else's. The 3dfx board manufacturer just tries to sell their boards to Voodoo zealots, who are, for the most part, GeForce believers now.

Tell me what makes you so afraid
Of all those people you say you hate

nope

[Jett]

The TNT was NOT the first consumer card with 32-bit support. That would be the Rendition Verite 2100/2200 (it also had a pretty decent full OpenGL ICD).

I agree 3dfx has lagged on implementing features, not counting the FSAA/T-Buffer deal in the VSA-100 generation which is pretty cute. It would be nice if they lead the market into new features rather than the other way around.

I strongly believe that 3dfx is positioning themselves for a solid comeback. They bought out Gigapixel which had some really great tech from everything I have read. Low power, high performance stuff, with very low transistor counts, they were a finalist in the X-box bid but lost out to Nvidia. The "Rampage" core has been in developement for a very long time now, a huge amount of 3dfx's R&D budget has gone to developing it. Considering the resources they've thrown at it I don't see any reason why it won't kick ass, unless they run into another component shortage (one of their biggest problems has been the RAM market).

I think 3dfx is going 1 of 2 ways: They either release the first "Rampage" core product and it kicks all ass and the company bounces back, or they release the "Rampage" product and it doesn't do very good and they continue their downward spiral and are bought out by another company. I personally think the first is more likely, but then I own a little 3dfx stock so I'm naturally a little optimistic.

The same mistakes, again and again

[TDSObeseWhale]

It is not too infrequent that we see many articles labeling gamers as "fanboys" of 3dfx or nVidia... Let me tell you right away, that 90% of gamers are not "fanboys". Like it or not, most gamers will become fans of the company with the fastest across the board chip, which currently is nVidia with their GeForce2 Ultra (ATI took the title for a bit, but that is arugable due to ATI's humiliating performance with Full-Screen Anti-Aliasing). Us gamers are not the strange breed many label us as, we don't tend to flock to a brand due to any sort of allegiance to a name, but rather we will end up flocking to the company with the best, fastest chip.

Now, onto my main point. 3dfx has a very thin fanbase, and there is a clear reason behind it. 3dfx, time after time, again and again, has been making the same mistake of denying the future. We've heard the argument that the Voodoo3's lack of 32 bit color support (and the memory to handle that feature) is what started 3dfx on its downward spiral. Yet, 3dfx has AGAIN made this mistake with their resistance twoards using a hardware transform and lighting solution... Any game developer will tell you that Hardware T&L is the way of the future, and 3dfx is shooting themselves in the foot. And what about the FSAA card? 3dfx did have a point when the voodoo5 was released and creamed the GeForce 2 in antialiasing performance, but those days are over, and with new drivers, the GeForce 2 beats the voodoo5 in its home territory, FSAA.

So to conclude, us "nVidia fanboys" have reason behind what we believe in. As soon as 3dfx comes out with a better chip than nVidia, count me in on the 3dfx bandwagon. -Matt "ObeseWhale" Grinshpun
-The Darker Sector
-Website coming soon! Team Corrosive Quake 3 mods.

There were advantages

[Bilestoad]

When the Voodoo II was current I was using a Matrox Millennium II with a Voodoo II - best of both worlds. The Matrox was an uncompromising professional card, laid out by engineers who understood analog and how to produce a sharp, sharp signal. The Voodoo was the best 3d around, of course.

Today you get all in one cards, but a lot of them are still not as good as the Matrox was. It's trial and (expensive) error to get one that's as satisfying in a text editor as it is in Tomb Raider. Anyone care to name a combined card, maybe a GeForce II or Radeon, that is as sharp as it should be in the highest resolutions?

The problem goes away when LCD monitors become affordable in 21/22" sizes, but I think that's at least a year away .

Turnarounds

[Fervent]

Given the relatively quick turnarounds technology companies have, I don't feel inclined to put 3DFX down for the count. They were a pioneer (along with PowerVR) and pioneers aren't always rewarded in their time.

In PowerVR's example, they no longer make PC cards in bulk, but their chipset is in use with the Dreamcast. It has shown to be surprisingly robust and has turned the Dreamcast from a dark horse into a system with some incredible games (Sonic Adventure, NFL 2K1, Jet Grind Radio, etc.)

Same goes for ATI, which has been running in neutral for the last few years, then released the jaw-dropping Radeon this year.

Point is, don't count 3DFX out yet. Their latest chipsets are nothing extraordinary, but a few engineers and some faith within their infrastructure might be enough to turn them around.

No surprise; most wounds are self-inflicted

[Froid]

Success generally breeds complacency, and even the most intelligent falter when they've finally reached the top and try to stay there. It takes a leader with a real inferiority complex (like Bill Gates's) to keep up the fight after all the competitors have been slain, and it takes a lot of cahones to squash all future competitors before they can rise to the challenge. Thankfully, many companies can recover after their pointy reckoning; look at IBM of today as compared to IBM of even ten years ago, and you'll see a company who's managed to extricate its head from its nether-regions after falling from the top. Will 3dfx be like IBM or will they follow the likes of Polaroid (and Kodak, at least with their copiers)? That remains to be seen.

We have a winner

[Reality Master 101]

That must be the most obnoxious article I've read this year. I think it actually has more "Gee-I'm-clever" lines (NOT) than actual content.

I read it, but I'm still not quite sure what it was about.

--