The 3500 seems to be the highest clocked voodoo card - 183mhz compared to 166 for the 3000, v4 and v5. So about 10% faster, probably only noticeable in benchmarks.
Anyone ever tested or benchmarked it against the other voodoos?
For all intents and purposes, a V3-3500 is equal ON PAPER to a pair of V2's. have a gander here: http://www.gpureview.com/show_cards.php I LOVE that site!
It is a mistake to think you can solve any major problems just with potatoes.
My V3-3000 will happily run at 3500 speeds with a fan blowing at it if you don't mind doing so.
wrote:For all intents and purposes, a V3-3500 is equal ON PAPER to a pair of V2's. have a gander here: http://www.gpureview.com/show_cards.php I LOVE that site!
On practice though, even a 3000 is faster than a pair of Voodoo 2 cards.
The 2000 is slightly less faster than a pair of Voodoo 2's in the worst case scenario!
I've had 2 or 3 3000 cards clocked at 183MHz and they were very stable. I'm pretty sure the TNT2 Ultra was faster than the 3500 though, especially in D3D.
Not sure about OpenGL!
Just search for any V3 with 6ns ram or better and they easily hit 183mhz on the core. The few v3 2ks out there with 5.5 and 5ns ram that are pci editions can crack 200mhz but they are very rare. Replace the crappy 10UF caps and you will get a little boost in max clock. The best that I know of is 216mhz for a v3. Personally the best bang for buck for those on agp is a v3 3k and on pci a 2k/3k with 5.5ns sgram or better. Those overclock like mad especially with a little extra cooling.
On a far away planet reading your posts in the year 10,191.
A V4 will be slightly faster than a V3 3500 when you run it using the same settings (16-bit color, low res textures, no AA) and have the advantage of being able to use a little bit of AA, 32-bit color, and higher resolution textures but at a cost in performance. A V5 is no comparison as it does everything faster than a 3500 even with all the bells and whistles turned on.
Some V3 2k and 3k boards actually used faster memory chips than what the boards were rated to run at so you can overclock some of them to reach the same speed as a 3500. Even some OEM V3 boards can get up to that speed but you need to add cooling because most of the OEM boards didn't even have a heatsink attached.
wrote:A V4 will be slightly faster than a V3 3500 when you run it using the same settings (16-bit color, low res textures, no AA) and have the advantage of being able to use a little bit of AA, 32-bit color, and higher resolution textures but at a cost in performance. A V5 is no comparison as it does everything faster than a 3500 even with all the bells and whistles turned on.
Some V3 2k and 3k boards actually used faster memory chips than what the boards were rated to run at so you can overclock some of them to reach the same speed as a 3500. Even some OEM V3 boards can get up to that speed but you need to add cooling because most of the OEM boards didn't even have a heatsink attached.
I wonder why they dropped use of SLI between Voodoo 2 and 5. There were plenty of faster derivatives of the Voodoo 2 in the Voodoo 3 series, but no use of SLI (either as chained PCI cards or multi-GPU boards a la Voodoo 5).
That probably could have kept 3DFX more competitive in the leading-edge high-end performance niche they'd originally established for themselves. (in addition to broadening their market with the more mainstream consumer/OEM oriented boards)
wrote:I wonder why they dropped use of SLI between Voodoo 2 and 5. There were plenty of faster derivatives of the Voodoo 2 in the Voodoo 3 series, but no use of SLI (either as chained PCI cards or multi-GPU boards a la Voodoo 5).
That probably could have kept 3DFX more competitive in the leading-edge high-end performance niche they'd originally established for themselves. (in addition to broadening their market with the more mainstream consumer/OEM oriented boards)
there are actually voodoo5 based sli cards, just not for gamers. the quantum3d alchemy machines use 2 pci66 cards sli, each with 8 vsa100 cores. 🤣
wrote:wrote:A V4 will be slightly faster than a V3 3500 when you run it using the same settings (16-bit color, low res textures, no AA) and have the advantage of being able to use a little bit of AA, 32-bit color, and higher resolution textures but at a cost in performance. A V5 is no comparison as it does everything faster than a 3500 even with all the bells and whistles turned on.
Some V3 2k and 3k boards actually used faster memory chips than what the boards were rated to run at so you can overclock some of them to reach the same speed as a 3500. Even some OEM V3 boards can get up to that speed but you need to add cooling because most of the OEM boards didn't even have a heatsink attached.
I wonder why they dropped use of SLI between Voodoo 2 and 5. There were plenty of faster derivatives of the Voodoo 2 in the Voodoo 3 series, but no use of SLI (either as chained PCI cards or multi-GPU boards a la Voodoo 5).
That probably could have kept 3DFX more competitive in the leading-edge high-end performance niche they'd originally established for themselves. (in addition to broadening their market with the more mainstream consumer/OEM oriented boards)
I suspect that has something to do with graphics cards moving to AGP (only 1 AGP slot per motherboard (there are exceptions)) and iirc AGP had difficulties with multi-GPU graphics cards
Well, they could have allowed you to SLI between AGP and PCI cards. 3dfx didn't use the AGP's advantages at all AFAIK 😜
It was probably their suicidal bad decision making and stuff.
What they needed to do was double down on Rampage and get that done a year earlier. Voodoo3 was a tweaked budget chip and VSA 100 was a stop gap because Rampage was going nowhere fast. Unfortunately, getting Rampage out when it was needed would probably have required major reworking of however they were running the company. Less feature creep or whatever was going on.
wrote:It was probably their suicidal bad decision making and stuff.
What they needed to do was double down on Rampage and get that done a year earlier. Voodoo3 was a tweaked budget chip and VSA 100 was a stop gap because Rampage was going nowhere fast. Unfortunately, getting Rampage out when it was needed would probably have required major reworking of however they were running the company. Less feature creep or whatever was going on.
In that context, having a Voodoo5-like implementation of the Voodoo 3 GPU series probably would have served as a much better stop-gap than the rather late release of the Voodoo 4 and 5, since they'd already lost a lot of market share by that point (and their top-end performance standing).
As it was, the Voodoo 4 and 5 GPU cores were only moderately better performers than the Voodoo 3 in general, and so late to market that they weren't compelling for mid-rage or high-end sectors. Whereas an SLI'd voodoo 3 (single and/or multi-board) derivative very well may have made a good high-end market leading product for the same high-performance niche as the Voodoo 1 and 2 has originally served.
Sure, if Rampage development was still lagging, the Voodoo 3 and SLI counterpart would have fallen behind by the time the Voodoo 4/5 were released in reality. However, unlike the 4/5, those cards could have been already established and much more popular (not to mention in high volume production), and perhaps aging at least somewhat gracefully into the mid-rage and low-end, leaving much less of a gap in the high-end market than having the Voodoo 3 alone (and late Voodoo 5) did.
wrote:I suspect that has something to do with graphics cards moving to AGP (only 1 AGP slot per motherboard (there are exceptions)) and iirc AGP had difficulties with multi-GPU graphics cards
The Voodoo 5 did just that though (dual GPUs on a single card), and wasn't the Rage Fury MAXX the only other dual-GPU card of that era at all? (albeit using AFR rather than SLI, and its problems weren't so much AGP specific as OS/driver specific)
On the multi-board end of things, they could have stuck to PCI or used mixed PCI/AGP with AGP-specific functions disabled. (if 3DFX cards even took advantage of that)
So, what I meant to say is 3DFX could have done with the Voodoo 3 what they later did with the 4 and 5 (in terms of lower-end single GPU models and high-end multi-GPU boards), if not continuing with the Voodoo 2's multi-board SLI support as well.
They had Rampage almost complete depending on ones point of view except there was no card that was close to being what they were planning for retail. Sure by being slow to evolve existing tech on hand and spending money as if were had rained down on them such as pissing away $50,000 plus a day on lavish employee lunches and no telling what other daily kickbacks before the party was over. As soon as the gravy train had derailed all else is history. They are the best example for any current and future company that being to confident of superiority can very easily lead to an early chapter 7/11 bankruptcy.
On a far away planet reading your posts in the year 10,191.
Apparently, the first Rampage cards were going to be released around April 2001. Or maybe that's just something I made up based on reading something else... Who knows really?
3Dfx was a trainwreck.
I remember reading that the separate T&L chip wasn't functional when they went under. That makes me think a final product was not even close. The other chips were seemingly operational but at an unknown performance level.
wrote:I remember reading that the separate T&L chip wasn't functional when they went under. That makes me think a final product was not even close. The other chips were seemingly operational but at an unknown performance level.
Rampage would have had to have been more integrated to be competitive with GeForce and Radeon. Releasing a card with that many chips on it would have made Rampage the first $600 mainstream video card and with a questionable performance benefit, if any benefit existed at all. Power consumption would have also been through the roof, worse than V5 6000.
Yup the multichip aspect felt wrong when their competition was fully integrated. Maybe it was a result of the long development cycle, or perhaps it made sense for Quantum3D hardware. They could scale geometry performance separately from the rest or remove it entirely.
wrote:wrote:I remember reading that the separate T&L chip wasn't functional when they went under. That makes me think a final product was not even close. The other chips were seemingly operational but at an unknown performance level.
Rampage would have had to have been more integrated to be competitive with GeForce and Radeon. Releasing a card with that many chips on it would have made Rampage the first $600 mainstream video card and with a questionable performance benefit, if any benefit existed at all. Power consumption would have also been through the roof, worse than V5 6000.
wrote:Yup the multichip aspect felt wrong when their competition was fully integrated. Maybe it was a result of the long development cycle, or perhaps it made sense for Quantum3D hardware. They could scale geometry performance separately from the rest or remove it entirely.
Yes, that's also what made the Voodoo 5 6000 impractical and the Voodoo 5 itself relatively cost ineffective with contemporary products.
At the time and with the relative cost/complexity of the Voodoo 2, the multi-board solution was more realistic (for the high-end gaming niche at least), as may have been for the voodoo 3 had 3DFX supported it, but by 2000/2001 the competing high-end performers made that difficult for the likes of the Voodoo 5, let alone more complex/expensive multi-chip designs.
OTOH, if it was only the T&L processor that was separate (with the main GPU fully integrated), and the design was relatively cost-effective otherwise (in terms of chip/silicon size, yields, power/cooling requirements, board size/complexity, etc), it may still have been a good idea. Designing it that way could also allow a modular set-up with low-end models omitting the T&L chip entirely, or allowing separate clock rates for the 2 chips for different performance models (especially important if yields/scalability varied significantly between the 2 chips).
With that said, it also might have made sense to release a single board dual-voodoo 2 card to appeal to the high-end without quite as high cost as 2 discrete voodoo 2 boards (or the inconvenience of taking up 2 PCI slots).
Granted, the original voodoo 2 was already a fairly large PCI board, so that might have been a tight fit even as a full-length card.
Sage was vertex shader 1.0 part with some other functions, so I would restrain from calling it a TnL chip. First batch already went to Nvidia's hands. The big hope of 3dfx was also Mojo. If there were money I would not write them off.
