On theretroweb, there's a PDF with some register information. https://theretroweb.com/chip/documentation/at … e4618791703.pdf

The ATI 3D Rate II is part of the "GT" group, in the "mach64CT Family". A section is called "Programming PLL Registers in mach64 CT Family". Mostly for MCLK/VCLK overclocking. There's also a ROM Header section with byte offsets.

--

On Bitsavers, a register reference guide for the VT and GT cards (Mach64CT family). https://bitsavers.org/components/ati/RRG-G027 … D_RAGE_1996.pdf

Everywhere it says "3D Rage" is for the GT (3d Rage I, II, II+, IIc, Pro).

Anybody know about the VT3, is it also part of VT4 and RageII or its more like older VT2 and has only directdraw support? I think its similar to VT2.

MikeSG wrote on 2026-08-29, 13:14:

On theretroweb, there's a PDF with some register information. https://theretroweb.com/chip/documentation/at … e4618791703.pdf […]
Show full quote

On theretroweb, there's a PDF with some register information. https://theretroweb.com/chip/documentation/at … e4618791703.pdf

The ATI 3D Rate II is part of the "GT" group, in the "mach64CT Family". A section is called "Programming PLL Registers in mach64 CT Family". Mostly for MCLK/VCLK overclocking. There's also a ROM Header section with byte offsets.

--

On Bitsavers, a register reference guide for the VT and GT cards (Mach64CT family). https://bitsavers.org/components/ati/RRG-G027 … D_RAGE_1996.pdf

Everywhere it says "3D Rage" is for the GT (3d Rage I, II, II+, IIc, Pro).

Thank you for sending them! But I already have both of these (RRG-G02700 Rev 0.10 and PRG-215R3)

RRG-G02700 is from 1996 and only covers the 264VT and the first 3D RAGE, so it doesn't have the registers that RAGE II / II+ / IIC added.
PRG-215 is for RAGE PRO.

What I'm still looking for is this one:
RAGE IIC and ATI-VT4 Register Reference Guide, Technical Reference Manual, RRG-G03600 Rev 1.01

If you or anyone has it, or any other document for 3D RAGE II / II+ (ATI-264GT-B), like a programmer's guide or a controller specification, please let me know.

Thanks again!

Hi Nortinette,

I don't have RRG-G03600 either. I hit the same gap writing a Windows 9x driver for the 3D Rage IIC AGP (1002:4757, 264GT2C), and I filled it by measuring the chip instead. It's all written up on GitHub, in case it's useful.

The register notes, put together from RRG-G02700, PRG-215R3, the 3D CIF SDK and the XFree86/atyfb/Mesa mach64 sources. Each claim cites its source, and anything I inferred is marked as inference:
ati-rage2-3d-engine.md

What I then measured on a real card (4 MB, Win98SE, Pentium III). Each write-up has the method and the raw probe logs:

• Register survey: register window, 2D engine, no setup engine
• Trapezoid edge model: how the engine steps a trapezoid, matched exactly
• Triangles and Gouraud
• Texture addressing: S/T units, sampling, wrap
• Filtering, texel formats, blending, fog, scissor
• Mip-mapping: how the engine picks a mip level
• Scanout start
• Triangle setup cost: the CPU-side setup a chip without a setup engine needs

The probe and scene tools are in the same repo under tools/diag (ati_rage_iic_probe*, ati_rage2_scene*), if you want to run them on your own card.

Caveats: all of this comes from one Rage IIC. I haven't checked any of it on a Rage II or II+ (264GT-B), and I haven't touched the VT4. Also, the PDF-to-text extraction of RRG-G02700 turned its bit diagrams into filler characters, so some bit positions in the notes come from the driver headers, not from ATI. If G03600 does turn up, I'd like to check these against it.

Only proof that this file existed is with this old post from Valery? https://www.os2museum.com/wp/ati-mach8mach32e … -documentation/
Some downloads here, but not exactly what you are looking : https://web.archive.org/web/20150108032518/ht … tories/doc/ati/

Some from archive.org
https://dn720701.ca.archive.org/0/items/bitsa … ide_1999410.pdf
https://ia801509.us.archive.org/1/items/bitsa … D_RAGE_1996.pdf
https://dn710603.ca.archive.org/0/items/bitsa … uide_199804.pdf

mwdmeyer wrote on 2026-10-04, 09:37:

Hi Nortinette, […]
Show full quote

Hi Nortinette,

I don't have RRG-G03600 either. I hit the same gap writing a Windows 9x driver for the 3D Rage IIC AGP (1002:4757, 264GT2C), and I filled it by measuring the chip instead. It's all written up on GitHub, in case it's useful.

The register notes, put together from RRG-G02700, PRG-215R3, the 3D CIF SDK and the XFree86/atyfb/Mesa mach64 sources. Each claim cites its source, and anything I inferred is marked as inference:
ati-rage2-3d-engine.md

What I then measured on a real card (4 MB, Win98SE, Pentium III). Each write-up has the method and the raw probe logs:

• Register survey: register window, 2D engine, no setup engine
• Trapezoid edge model: how the engine steps a trapezoid, matched exactly
• Triangles and Gouraud
• Texture addressing: S/T units, sampling, wrap
• Filtering, texel formats, blending, fog, scissor
• Mip-mapping: how the engine picks a mip level
• Scanout start
• Triangle setup cost: the CPU-side setup a chip without a setup engine needs

The probe and scene tools are in the same repo under tools/diag (ati_rage_iic_probe*, ati_rage2_scene*), if you want to run them on your own card.

Caveats: all of this comes from one Rage IIC. I haven't checked any of it on a Rage II or II+ (264GT-B), and I haven't touched the VT4. Also, the PDF-to-text extraction of RRG-G02700 turned its bit diagrams into filler characters, so some bit positions in the notes come from the driver headers, not from ATI. If G03600 does turn up, I'd like to check these against it.

Thank you very much, this is useful. I am currently emulating the Rage II+ and have never had an actual card for comparison, so your measurements provide the first set of data from real hardware that I’ve had access to.

The emulation already matches the measured behavior in several areas such as edge walking, pixel spans, fill direction, Z-testing, and texture scaling and wrapping but there are a few slight differences:

Color overflow (saturation from the 256–383 range down to 255)
Mip level threshold
3-bit bilinear weights
S/T precision

Initially, I understood that the Rage IIc would likely perform similarly, though I also suspected there would be some modifications compared to the Rage II+
Since your card is an IIc model, I am hesitant to modify II+, I am guess that both chips operate almost identically.

marxveix wrote on 2026-10-04, 15:31:

Only proof that this file existed is with this old post from Valery? https://www.os2museum.com/wp/ati-mach8mach32e … -documentat […]
Show full quote

Only proof that this file existed is with this old post from Valery? https://www.os2museum.com/wp/ati-mach8mach32e … -documentation/
Some downloads here, but not exactly what you are looking : https://web.archive.org/web/20150108032518/ht … tories/doc/ati/

Some from archive.org
https://dn720701.ca.archive.org/0/items/bitsa … ide_1999410.pdf
https://ia801509.us.archive.org/1/items/bitsa … D_RAGE_1996.pdf
https://dn710603.ca.archive.org/0/items/bitsa … uide_199804.pdf

Most of the documents I have now are from online websites. also that you provided, thank for help.