3Dfx Rampage testing (2013)

gdonovan

[H]ard|Gawd
Joined
Oct 7, 2004
Messages
1,818
DSCN5649.JPG


Good morning [H]orde!

I made mention of this in one of the board threads a few weeks ago and today I have a few moments to make a post. Without going into too much detail or history I'll simply state I have one of the few fully functioning (to some degree, more later) 3dfx Rampage boards which I have owned since 2005 and last year finally got around to doing some work when a German 3dfx enthusiast stepped forward to make a prototype dongle to invert the output.

Lets get some misconceptions out of the way first.

1) There are 2 fully operational Rampage boards floating about and three known "clamshell" boards. There is also one clamshell board with half the ram that was used for testing. 3dfx made roughly 10 boards the 46th week of 2000 and 10 boards the next week. There is a confirmed number of 24 chips made, I know where there is at least 4 if not more loose chips.

2) The reason for the dongle is the chip was modeled in software first and the DAC output was overlooked. The dongle flips the output.

3) The only (current) drivers are very alpha win9x which I have tried under Win95A, Win98SE and WinME. The drivers work but only in a limited sense. The engineers were trying to focus on working on the chips core aspects and some aspects of memory management were disabled, were working on this.

4) Rampage was a DX8 part, it had very little in common with Voodoo Graphics --> VSA-100. It was to be a clean sheet design, native DirectX part with native D3D/OpenGL with Glide supported via a wrapper.

5) Rampage is very picky about what motherboard it will run with, these has caused me some headaches getting everything operating.

6) I have a limited number of DirectX demos and games running and a few OpenGL games running with a wrapper. Some applications will run for some time, others will crash out in a few seconds.

7) Since this was a first batch chip there was a lot of stuff disabled so they could do debugging, one of them was the memory interface as well as most AGP functions. Once I thought of the card as a dumb PCI device I had more success getting it to run reliably. 3dfx was debugging the board & chips and working on the drivers right up till the moment the doors closed. Too late as the rampage boards were still a year from commercial sale, not enough to save the company.

Just to be clear, I have no illusions that the board will ever run to its full potential but there are a number of people (including other Rampage owners) would would like to see them up and running as best they can. A few of the ex-3dfx people have been helping out and there are a large number of enthusiasts who have chipped in as well.

Next I'll post some pictures of the various boards and links to videos.

Gary Donovan

P.S. Some information here- http://www.thedodgegarage.com/3dfx/rampage.htm

And last years fun and games here- http://www.thedodgegarage.com/3dfx/rampage_2012.htm
 
Last edited:
Ah, for the days when a gpu cooler didn't need planning permission to install it in your case :(

3dfx management really needed a good kick in the balls, far too many stop gap "filler" cards instead of focusing on the (then) next gen stuff :(

Still find it funny that rampage board has a few caps at the power connector (oversight), though if I remember right it didn't need the connector anyway?
 
image-o-matic1.jpg


image-o-matic2.jpg


These two images are from 3dfx that leaked out showing them bringing up the 16mb board and working on the ram and BIOS.

801161SpecterA0front.jpg


Here is that exact board today, yes it still survives.

rampy_juntas.jpg


Two socketed boards, one works and the other does not. Currently the non-functioning one is undergoing rework to get it up to speed. One of the reworks was to do away with the clamshell socket and directly attach the Rampage chip to the PCB which is done at this point.

rampagedongle.jpg


This was the first Rampage board to surface several years ago.

specter_dongle.jpg


A close up of the original dongle, the board and owner has been most helpful in the creation of the duplicate I'm using.

reva2_1_1fzu8a.jpg


The duplicate I'm using at the moment, its fabricator is currently working on the next version which will have a better picture.

I know where there might be one or two other rampages but better to leave that discussion for another time.

a0_front.jpg


Oh, and here is a blank PCB.

Gary
 
Still find it funny that rampage board has a few caps at the power connector (oversight), though if I remember right it didn't need the connector anyway?

Correct, since it was modeled in software (a first) they were not sure if the AGP connector would provide enough power. So when the board was populated it was just left on, would have been done away with on the next revision.
 
DSCN5889.JPG


What output looks like without dongle.

DSCN6033.JPG


Output with dongle, Everest properties page.

newpage1.jpg


Some boring 3dfx properties, I'd take some of that with a grain of salt. The BIOS revision is correct but the board has 32mb of DDR ram.

hl_direct3d.jpg


Half-life in Direct 3D

quake04.jpg


GLQuake with OpenGL wrapper.

q3_simple_2.jpg


Quake3 running with a simple config before stepping up the graphics options.
 
Psst... [H]orde.

Very cool stuff! I was rocking a Voodoo2 12MB + TnT setup for a few years until about '99 when I got a GeForce2. Love me some 3Dfx. Good times, good memories.
 
Often wonder how things would have progressed if Rampage had been out on time and got 3dfx some momentum. Would ati be as big of a player as they are today? Would it be a three horse race instead of a 2 horse race? Oh well =/
 
I certainly remember the Rampage being right around the corner when they folded.

I wonder what the box art would have looked like. :p
 
That is cool stuff. Thanks for sharing. I will be coming back here to continue to saga as it unfolds! :)

3dfx was awesome in it's prime. I still have a couple Voodoo II's somewhere (SLI). I didn't get them until 3dfx folded. I had bought a Matrox m3D card (for $150 then a couple months later they were discounted to $15). But, I did have the satisfaction of being able to play on them. Nice to see some unreleased hardware, and glad there are still some resources out there to help you out.
 
wow

I wonder if any of the original engineers know this is going on
 
Running System Shock 2 demo-

https://www.youtube.com/watch?v=zAhWPBqVqac

Running Quake3-

https://www.youtube.com/watch?v=auAqCV0vK9s

Running Final Reality benchmark-

https://www.youtube.com/watch?v=7OzhMd2q6Zk

Running DX 7, 8, 9 diag tests-

https://www.youtube.com/watch?v=kmTSRdE-Gb4

Running a bunch of old DirectX tests with a V4-4500 next to it for a point of comparision.

https://www.youtube.com/watch?v=NomMCmGdHFk

I'm just gunna go ahead and quietly make sweet love with this entire thread. I'm not even going to ask for permission...
 
I certainly remember the Rampage being right around the corner when they folded.
It looked like it was still in development, an all hands on deck priority, when 3dfx folded. The company could never get anything out on time. When 3dfx died, Rampage was almost a year behind schedule (2 year project started in early 1998) and hadn't reached final silicon in late 2000.

Wishful thinking aside, it wasn't close to shipping, or as advanced as some people still like to claim. If it had shipped in early 2001, it would have competed against the GF3. If it had shipped in mid 2001, it would have competed against the ATI R200 (Radeon 8500). Both cards had equal or better specs in every way than 1 Rampage + 1 Sage, and likely would have been much cheaper than Sage equipped 3dfx cards.

Even if the card the OP posted came out before 3dfx ceased development of products, the GF2 and Radeon DDR were already out months earlier with many of the performance targets (except of course the amazing Sage chip which never had to face actual performance tests) that Spectre was supposed to hit.
 
It looked like it was still in development, an all hands on deck priority, when 3dfx folded.

No, there was other projects going on.

VSA-101 (Daytona) the follow up to VSA-100 was on track for release in 2001, indeed 3dfx even sold a pile of Daytona boards to IBM! VSA-100 with a die shrink and DDR support.

Plans for single chip, dual chip and quad chip boards were in the plan and Daytona overclocks like mad, 250 mhz with stock cooling. For you new comers, VSA-100 stock was 166 mhz and usally run 183 without much issue with 200-205 being the utter max with cooling and a overvolt.

Wishful thinking aside, it wasn't close to shipping, or as advanced as some people still like to claim.

As I stated above, point number seven and it was pretty advanced for its time. Feature creep was one issue that delayed the chip as more stuff was piled on.

If it had shipped in early 2001, it would have competed against the GF3.

Dual Rampage with SAGE would have had GF3 for lunch.

But neither here nor there, rehashing history 12-13 years after the fact won't change nothing.
 
Dual Rampage with SAGE would have had GF3 for lunch.
Santa also delivers millions of presents a night one day a year too. ;)

The glowing revisionism about 3dfx just gets annoying. It was a poorly run company with products that rarely had advantages vs the competition (Glide, while it lasted, aside).

Don't get me wrong. I liked 3dfx cards way back when. I had Voodoo Graphics, Voodoo 2 (and V2 SLI), and a couple of Voodoo 3 cards. The company was just falling farther and farther behind and was not some kind of leading edge GPU maker, despite how loudly fans tried to state otherwise.
 
Santa also delivers millions of presents a night one day a year too. ;)

The glowing revisionism about 3dfx just gets annoying. It was a poorly run company with products that rarely had advantages vs the competition (Glide, while it lasted, aside).

Don't get me wrong. I liked 3dfx cards way back when. I had Voodoo Graphics, Voodoo 2 (and V2 SLI), and a couple of Voodoo 3 cards. The company was just falling farther and farther behind and was not some kind of leading edge GPU maker, despite how loudly fans tried to state otherwise.

Well the biggest problem 3Dfx faced was trying to maintain some sort of monoploy by perpetuating their Glide API when everyone else was able to quickly develop hardware for OpenGL and DX features. 3Dfx put most of their eggs in the wrong basket...ignoring rapid tiered/segmented hardware development for basic industry standards and mass quantity manufacturing plus securing said mass quantity for deals with OEMs.
 
The glowing revisionism about 3dfx just gets annoying. It was a poorly run company with products that rarely had advantages vs the competition (Glide, while it lasted, aside).
Granted, but I think most of us look back fondly on 3Dfx for really bringing us into the era of 3D acceleration. It may've only been bilinearly-filtered, limited to a 16-bit palette and to 640x480 (or 800x600 if you didn't use the depth buffer), but it was still an order of magnitude better than what we were dealing with before. You could certainly get 3D acceleration before the Voodoo, but not for $199 or $249 or whatever those sold at.
 
The glowing revisionism about 3dfx just gets annoying. It was a poorly run company with products that rarely had advantages vs the competition (Glide, while it lasted, aside).

Revising history? You are perhaps, but I have actually benchmarked a Geforce3 vs a V5-6000 so have some frame of reference for which I speak. Even with "only" a V5-6000 the battle between the two was largely a wash with GF3 have an edge in D3D and 3dfx carrying the day in glide/opengl applications.

A dual Rampage board would have had double the memory bandwidth of a GF3-200 and stil have more than 2.2 GB/S over the top dog GF3-500. Dual chip graphics carry large advantages, a path followed to this day eh?

Again, there is no point rehashing history.
 
Granted, but I think most of us look back fondly on 3Dfx for really bringing us into the era of 3D acceleration. It may've only been bilinearly-filtered, limited to a 16-bit palette and to 640x480 (or 800x600 if you didn't use the depth buffer), but it was still an order of magnitude better than what we were dealing with before. You could certainly get 3D acceleration before the Voodoo, but not for $199 or $249 or whatever those sold at.

Kids have no idea how good they have it... perhaps we could chain one to a Pentium 90 machine with DOS and a 1mb cirrus logic video board and watch the hilarity ensue.

Now get off my lawn.
 
perhaps we could chain one to a Pentium 90 machine with DOS and a 1mb cirrus logic video board and watch the hilarity ensue.

Need input: ISA, EISA, VLB, or PCI?

Of course, mentioning Pentium 90, probably either VLB or PCI.
 
Want to thank board member Kuschelweich for sending me a Abit KT7A socket A (KT-133A) motherboard to try out, unfortunatly its not playing well with Rampy and will crash out after a few seconds under D3D operation.

So far thats 3 BX boards, one KT-226, one LX/EX that have failed to work (to one degree or other) so far the runtime champ is a VA6 with ApolloPro133 chipset with a Epox KT-133 board nipping at its heels.
 
Holy shit this thread brings back some dear memories over a decade ago. Still have one of those early S3 Virge and Matrox Mystique cards lying around.
 
What an amazing piece of work! Thanks for doing this and keep it up Gary!
I still remember reading about Rampage well before I purchased my Gf3 Ti 200, oh to speculate what could/would have been....
It makes me wonder as well if 3dfx would have "fixed" all the frametime multi-GPU hitching that has been around for many years well before Nvidia/Ati.
Never knew about VSA-101/Daytona though, if you have any other tidbits of info keep em coming!
I obviously have nothing better to do at 3am.... :eek:
 
Need input: ISA, EISA, VLB, or PCI?

Of course, mentioning Pentium 90, probably either VLB or PCI.

PCI...by the time the 90MHz Pentium came out, VLB was pretty much gone from the Pentium motherboards. VLB had some compatibility problems with the memory bus on the Pentiums as they were designed to go hand in hand with the 486. There were only a few motherboard makers who had the combo boards via a third party chipset.
 
Last edited:
Granted, but I think most of us look back fondly on 3Dfx for really bringing us into the era of 3D acceleration. It may've only been bilinearly-filtered, limited to a 16-bit palette and to 640x480 (or 800x600 if you didn't use the depth buffer), but it was still an order of magnitude better than what we were dealing with before. You could certainly get 3D acceleration before the Voodoo, but not for $199 or $249 or whatever those sold at.


i remember how i went to the neighboring city only because someone told me about that "wonder card" that made quake look so totally different... i went, saw quake running in 640x480 at literally twice the frame rate (28 vs 14 fps, lol) and came back with the card. that was like the moon landing for games...
 
i remember how i went to the neighboring city only because someone told me about that "wonder card" that made quake look so totally different... i went, saw quake running in 640x480 at literally twice the frame rate (28 vs 14 fps, lol) and came back with the card. that was like the moon landing for games...

Pretty much the same here- A fellow gamer has a Micron 200 mhz system and we were playing "one up" on upgrades. My system was a touch faster running a home built Asus with 75 mhz bus but he went out and purchased a Orchid voodoo graphics.

I had a monster board the next week ($400), glquake was awesome.

Still have it to along with the box and discs.

I actually still have the system I built after that- Pentium MMX 233 running at 266 mhz with an ET-6000 for the primary video and a Sound Blaster 16.

DSCN3143.JPG


Forgot about the 3-com NIC carb with coax and 56k modem.

DSCN3136.JPG
 
Last edited:
Nice thread! Loved 3dfx myself.. those were the good old days were you could buy a graphics card and sit on it for 2 years before the need to upgrade!

I've had plenty of 3Dfx cards in my time and sold most of them in the original boxes a few years ago and then built some more old systems and snagged up the cards again hehe.. Voodoo2 12MB cards in SLI was awesome.

Sad thing is that i'm basicly being forced to sell off my old pc parts soon as I am testing stuff and getting it all together to see all that I have.. I try to tell her it aint worth much but she don't care.. If you can get $20 she will take it! :/


Makes you wonder what today would be like if 3Dfx was still kicking.. I bet they would be showing these other guys how to run multipal cards the right way!

Good luck finding that magical mb to run that rampage!

I do have a few older mbs but no idea off the top of my head.. Some need some new caps before they will run.
 
rampage+screen+trash.jpg


Type one failure.

DSCN6946.jpg


Type two failure.

Motherboard incompatibility so far run into three catagories- Garbled display on boot up (1) display artifacts (menu ghosting) after the driver is loaded and in 16/32 bit (2) and the last is absence of the first two failures but crashing shortly when most 3d applications are launched. (3)

BIOS SETTINGS.

Set all video shadowing and caching to off, Set AGP to 1x, Shut USB support OFF and most of all set AGP aperture size as low as possible! Most motherboards go only as low as 4mb. Set all AGP related read/write settings to off.


dx9page.jpg



In the Windows DXdiag panel set AGP Texture Acceleration to disabled.

Best OS to date has been Win98SE with a slight edge over WinME, running DX9.0B.

Do NOT install any VIA chipset packs, I tried 4.35 and it caused massive problems.

1) Best motherboard so far (by a longshot) has been an Abit VA6 (ApolloPro133 chipset) running a P3-800 in a slocket adapter, BIOS date 9/26/00. USB on or off seems to make no difference.

2) Next best has been Epox 8K3A+ (KT-333 chipset) if USB is left on in BIOS you will get error number two shown above. BIOS date 9/16/03

3) Abit KT7A (KT-133A) Version 1.3 with BIOS date 8/22/02 has type three failure, tried both WinME and Win98SE with little to no change.

4) Abit BE6 (Intel BX chipset) Version 1.1, rendering error two, unable to disable USB.

5) Epox 8KHA+ (KT-266A chipset) error number one.

6) Biostar? MT6BA (Intel BX chipset) Version 1.5, error number one.

7) Tyan Dual slot one motherboard with BX chipset, failure two.

8) Compaq OEM motherboard with LX/EX chipset, failure two.
 
Interesting...

You would think the ones with the BX chipset would be the more stable.

Any ideas why?
 
Interesting...

You would think the ones with the BX chipset would be the more stable.

Any ideas why?

The BX was a popular platform (indeed you can see ram and video bios work being done on one in the first picture), the only thing I can think of was that the 3dfx engineers were using something different at the time or the BX motherboard shown in the picture had special BIOS options available.

BIOS of the motherboard seems to factor heavy on video function, why does the KT-333 work somewhat but the KT-226A does not even though both are Epox motherboards? Some BX boards almost work while one had nothing but garbled trash.

There were bugs to be worked out of the Rampage boards BIOS, they just ran out of time. My best bet is to find out what platform was being used when they were working on the drivers and emulate it. So far no one can give me an answer or remember. One rumor was they were doing work on a Slot A which was a rather powerful platform at the time and the board does seem to favor VIA chipsets.
 
Back
Top