[EDIT]hmmm.....whats going on at AMD? (and history of 3Dfx Rampage)

[ocre]

OpenGL was always Nvidia's strength.  Even TNT2 Ultra gave Voodoo3 3500 a serious run for its money in Quake3, in 16-bit color mode.  Voodoo5 6000 still came within a few % of Geforce2 GTS in Quake3, at 1280x1024 resolution.  The scaling was wonderful for that resolution.

It was quite simple to do SLI back then.  Surely, it would have been ready as well, if 3dfx didn't have all of these financial problems.  Obviously, R&D was being cut, projects were being delayed, etc..   But then again, putting 2 of these chips on a card was something 3dfx engineers could easily do, as proven with the awesome Voodoo5 5500 that I had (awesome for full AA in games just a couple years old like Need for Speed 3, etc..).  The Rampage Specter 3000 would've mopped the floor with Geforce 2 Ultra, the first $500 gaming card.  With DX8 and additional DX9 capabilities, even Geforce3 would've had a hard time keeping up with the dual Specter with Sage T&L unit. 

The voodoo5 5500 was two VSA-100 chips on one PCB, the vsa-100 was an updated avenger chip used in the voodoo3.  In a time when 3dfx was really slipping behind, they spent valuable time updating the avenger architecture to finally release the dual chip voodoo5 5500 in june of 2000. You may or may not know this but one of the major reasons in updating the avenger chip was scalability, its design centered around using more than one chip on a single board. The letters in VSA-100 stand for Voodoo Scalable Architecture.  

But why waste valuable time updating the Avenger architecture ballooning its 8 million transistors to 14 million if it was just so easy to put 2 chips on a single card all along.  Why didnt they just throw 2 avenger chips on a single pcb and had a card nearly as powerful as the voodoo5 5500 -a year and 3 months earlier- in march of 1999?    

3DFX was struggling to keep up and their solution was to use multiple chips in order to compete.  The VSA-100 came as a result of that ambition, the first GPU to use the chip was the Voodoo 5500, which was a dual chip design.  The only issue in my opinion is that 3DFX didnt have the valuable time it took to make that happen.

The Spectre 3000 never existed, not even in a C-clamped and duct tape form.  We can only imagine if the Rampage already had VSA-100 like scalablity features baked in its design....but if it did, then what a waste for engineers to be sampling and testing single chip configurations.  Especially more interesting considering that the VSA-100 debuted in a dual chip GPU right off the bat.  And i guess that is one of the bigger issues i took with the blog and information you first posted.  The figures and numbers the guy spouts out ("10.2 GB/S bandwidth with a fillrate of 1600-2000Mpixels/S. To give you a point of reference Nvidia didn't have anything matching these performance specs till....") and the story of how they had everything done but were laid off before it could be mass produced, they dont align at all.

the only thing engineers had was some very very early stage single chip prototypes. So quoting specs of the 100% imagined spectre 3000.........there was none of those ready.  The feasibility and ability of the rampage is a complete unknown, the 8 or so engineering prototypes dont tell us much except that they were in development of a new architecture.  But there is a lot that would have had to happen from that point until the spectre 3000 could have launched.  For starters, engineering samples which just dont exist.    

Apparently, you missed this part:

Although there's only 1 TMU per pipe, the chip can out match the NV20 in Multi-texturing as it can apply 8 textures in a single pass. This is done using Rampage's texture computer and its unique "loopback" function.

And this part as well:  "The REV-A1 boards screamed. We were laid off on Friday we had debugged that board till Thursday night. It ran Quake3, It would have been a real threat to Nvidia due to cost and performance. We hit 250 Mhz DDR"

Why would they lie about it, long after it's been history, long after 3dfx was out of business??  Anyway, the blogger does not claim that Rampage would've saved the company - being humble enough to say

"Could Rampage have saved the company? Without knowing all the facts it is tough to say, but it would have been another 6 months before the first SPECTER boards started showing up on the shelves with only the Daytona and the older Napalm chips to shoulder the burden till then. The Specter family of boards was very powerful and would have outperformed any other current graphics solutions by a wide margin."

So it ran Quake? okay, that is great. But how well? Yeah......that is kind of important.
Why would they lie? That is interesting cause i am not calling them liars. Just saying.....
There is no way that anyone knows what would of or could of happened. We do know for sure what did happen though. And anyone claiming this or that would have been the outcome in some made up event that is all in there head........that is all just wild speculation.

If he says they were able to run the DDR at 250mhz, well that is great!!!!
But is that supposed to convince me that nvidia would have been in some serious trouble? HOW?

Do you remember when intel was wiping the floor of AMD using bulky and much slower front side buses? The 250mhz DDR doesnt somehow mean that nvidia would be in trouble nor does the fact that the prototype board actually booted up and ran quake. I mean, at the very least......shouldnt it boot up and play at least one game? And that is all we have, some paper specs and the very least real info we could possibly have. It booted a game, WOW!!!

Well, 3dfx was already doomed with the delayed Voodoo5 anyway.  If they never bought STB, it would have been delivered at least 6 months earlier, preventing some more of the damage that Nvidia was doing with their full 32-bit color support (with a heavy 50% performance hit that was usually not feasible until the DDR version of Geforce), and their heavily advertised T&L unit that was not practical until Max Payne a couple years later (when Geforce3 was out anyway).  3dfx would've continued to work hard on their projects, perhaps even hiring more engineers, and delivered Rampage before Nvidia's Geforce3. 

But Ocre, maybe you're right - like the plagued Voodoo5 6000 that was never really fixed, perhaps 3dfx would've also had issues with Rampage in SLI and with that T&L unit.  It would've been kickass to have RGMSAA and 16x AF long before Nvidia or ATI made these features feasible.  Just imagine how much better games would've looked with at least 8x AF with its efficient 8-pass "loopback" texturing function?

Also, I held onto my Voodoo5 5500 for years, for optimal Glide experience with key games such as Unreal Tournament, Half Life, etc..  Couldn't stand the glide wrappers that not even my Geforce4 Ti could handle as nicely.  Glide might have lived on a bit more if Voodoo5 wasn't delayed, and especially if Rampage was produced. 

Ahh, nostalgia during my most crazed gaming period of my life!  When playing games so much, for hours and hours, more IQ could not be overvalued.

[/quote]

The Rampage could have been all that and a bag of chips, it could have been the excalibur of GPUs -OR- it might have not have been able to compete at all.
Just saying it had DDR or it ran at 250mhz, it is all pretty meaningless. Using it is one thing, how well it uses it is another. And this, we know nothing at all about.

I am just saying, the only chip that we see was being worked on is a single chip prototype. The specs of a single Rampage arent impressive at all, not when you look at other chips in the same era. And for all the special claims, the only real thing that matters is the real world performance of the final product, which was a good ways off. The chip might have needed a respin or erratum anything. There might have been issues in their new DDR memory controller.......just look at this for a minute. These were the very first samples of a new design that just got in the engineers hands. They were largely untested and very little known when it comes to their actual performance. I mean, my point is.......All we know about it is the hype. That is it




edit--------
i cant stop chuckling when i think of the DDR "250mhz" statement. I mean surely bofox, you know that an individual max capability of a prototype chip and the actual feasibility (when it comes to mass producing) are two drastically different things. Binning, binning, binning. I am sure nvidia has gm200 chips that could reach well over 1600mhz in the lab....do i really have to go any further? Anyway, just thought i might add that.