Cards that I want

These are the cards that I'd like to get for X / DRI hacking. Within each group they are sorted from most wanted to least wanted. Some of the orderings will surprise people, I'm sure. As with my list of cards that I have, the XML and stylesheet are also available.

3dfx

There's pretty much no maintainer for the Voodoo3 DRI driver, and that sucks. The PCI versions are nice because you can put them in a wider variety of systems. Old AGP cards tend to not work in modern AGP 8x systems. This card is #2 on my 3dfx list because there are so many of them out there. I think the Voodoo3 and Voodoo2 represent 90% of the 3dfx cards in the wild.

So, I've still got a soft spot in my heart for the old 3dfx chips. I sometimes wish I still had my old Canopus Pure3D. The SLI and dual TMU Voodoo Graphics cards made by Quantum3D are also pretty interesting. Of course, they're getting increasingly rare and expensive. At some point they will cost more than they did new. Realistically, I think the SB50-4440 or SB50-2440 are the ones I'd most like to have.

3dlabs

I have acquired a Permedia2 card for x86. There are also Permedia2 boards available for PowerMacs, Sun workstations, and, believe it or not, Amigas. I wouldn't mind picking up one for a non-x86 architecture.

Someone recently reminded me that this card existed. It's one of the oldest game oriented 3D accelerators (released in 1995). As far as I know, it's the only 3D accelerator available for VESA local bus (VLB). As near as I can tell, the GiGi is a scaled-down (believe it or not) version of the 300SX.

ATI

As large as my collection of ATI cards is, there are still a few useful (well...) or interesting cards that are missing.

Did I mention that I like quirky hardware? I'm not sure hardware gets much quirkier (is that a word?) than this. This was the card that ATI hoped would keep consumers happy until the Radeon was available. I don't know if it made consumers happy or not, but I'm pretty sure it made driver writers miserable! Instead of doing SLI like 3dfx, ATI configured each chip to render alternate frames. This, of course, only really works if you have only one 3D window open at a time. In any case, the performance boost wasn't that great, and the drivers had lots of problems. Knowing what I now know about driver development, I'm surprised their driver team didn't go on strike when they were given this card!

Why do I give a rats ass about this card? Well, the open-source Rage128 driver ought to be able to support using this as a normal, single-chip card. However, there seem to be some problems, and it just doesn't work. That bothers me.

This is one of ATI's oldest 3D cards. It's so old, in fact, that I'm not even sure that Quake 3 runs on it without disabling some features. This is one of those first generation cards that did not do triangle setup in hardware. Triangle scan conversion is done in software, and the parameters (e.g., screen position, colors, texture coordinates, etc.) for each scanline are sent to the hardware. These types of cards are very slow because the CPU has to do a lot of extra work and they have to send a lot of extra data over the bus.

This card is "interesting" because it was used in the first generation iMac. It was also used in a few notebooks, and I think it was even used in add-on boards for some off-beat platforms like the Atari Falcon and Amiga. In terms of making bottom-of-the-barrel old hardware more usable in Linux, it would be nice to have at least some 3D support for this card.

Matrox

I recently acquired an AGP G200. However, I'd still like to get a PCI card. PCI and AGP cards, particularly for texture uplloads, are treated differently. It would be nice to test both.

Anyway, there are several flavors of the G200 available. Most of the earlier ones shipped with 8MB installed and a SO-DIMM slot for an additional 8MB. Some of the later ones shipped with 16MB on board. IBM shipped a PCI G200 with some pSeries systems called "GXT130P".

Rendition

I've acquired a 4MB PCI card (V2100), but an 8MB AGP card (V2200) would still be somewhat interesting. The Verite 2x00 chips (the V2200 is the AGP version, and the V2100 is the PCI version) are based around a highly programmable RISC core. I don't think it's quite programmable enough to do fragment programs or anything like that, but I'm sure an open-source driver could support more functionality than the Windows drivers ever did.

I'd love to get my hands on a Hercules Thriller Conspiracy. These cards added a Fujitsu FGX-1 (Pinolite) (an english translation by Google is available) geometry accelerator to the mix. That would have made it the first consumer card with hardware TNL. To be perfectly honest, I'm not sure if any of these cards ever made it into the wild. I've never been able to find a picture of one on-line, and I've never seen one listed on eBay.

Micron (who bought Rendition) has a semi-complete list of boards based on all the Verite chips.

S3

I know, I already have a crappy Virge card, but this one is different! This is a Zorro II card for the Amiga. As far as I know, this is the only Zorro card made for the Amiga that has any 3D acceleration. Most of the other Zorro graphics cards were either based on Cirrus Logic chips or quirky, custom 2D-only designs.

The Vivid 24 could also do 3D accelation, but that card is too quirky even for me!

SGI

This is the integrated graphics in SGI's 320 and 540 Visual Workstations. I guess the graphics core is almost identical to the one in the O2, but that's just hearsay.

Revision History

Removed entry for Revolution IV because I got one.

Removed the PCI-e G550 entry because I got a card.

Added entry for GameGlint card.

Updated the Rendition entry after acquiring a "Stealth S220" on Monday.

Removed Banshee from the list (because I got one). Updated the G200 entry (because I got an 8MB AGP card).

Added info about a bunch of cards on the list.

Added link to the "cards I have" page.
Added links to the raw XML and style-sheet.

Sorted companies by name (instead of randomly).
Added "Rage Fury MAXX" and "Rage IIc" to the card list.

Initial version.