Matrox G200
Matrox's first fully AGP-compliant graphics processor with 32-bit color.
The Matrox G200, released in 1998, was a graphics chip for PCs that handled 2D, 3D, and video acceleration. It aimed to merge Matrox’s strong 2D and video performance with a fully capable 3D engine. The chip was used on several boards, most notably the Millennium G200 and Mystique G200. The Millennium G200 came with SGRAM memory and a faster RAMDAC, while the cheaper Mystique G200 used slower SDRAM but included a TV-out port. Most G200 boards shipped with 8 MB of RAM, expandable to 16 MB via an add-on module, and also had ports for optional boards like the Rainbow Runner, which added extra features.
The G200 was Matrox’s first graphics processor to fully support the AGP bus. Unlike the earlier Millennium II, which was adapted for AGP but lacked the full feature set, the G200 used Direct Memory Execute (DIME) to speed texture transfers to and from system RAM, allowing it to use system memory as texture storage when the card’s local RAM was insufficient. The chip had a 128-bit core organized as two separate 64-bit unidirectional buses, which Matrox called “DualBus.” This design reduced data transfer latencies by improving bus efficiency compared to a single wider bus. The memory interface was 64-bit.
The G200 supported full 32-bit color depth rendering, which improved image quality by eliminating dithering artifacts common with 16-bit color. Matrox branded this as Vibrant Color Quality (VCQ). The chip also supported trilinear mip-map filtering and anti-aliasing, though the latter was rarely used. It could render 3D at any resolution supported in 2D. Architecturally, it had a single pixel pipeline with one texture management unit, and a RISC processor called the “WARP core” handled triangle setup in microcode. The G200 was Matrox’s first graphics processor to require a heatsink for cooling.
In 2D, the G200 was fast and delivered Matrox’s well-regarded analog signal quality, outperforming the older Millennium II in most areas except very high resolutions. In 3D, its Direct3D performance was similar to but generally behind a single Voodoo2, and slower than the NVIDIA Riva TNT and S3 Savage 3D, though it remained competitive. Its 3D image quality was considered among the best, thanks to 32-bit color support, provided driver bugs were not an issue.
The G200’s main weakness was OpenGL support.
- Release year
- 1998
- Core bit width
- 128-bit (DualBus organization with two 64-bit unidirectional buses)
- Memory interface
- 64-bit
- Standard ram
- 8 MB, expandable to 16 MB
- Color depth
- 32-bit (Vibrant Color Quality)
- Manufacturing process (original)
- 350 nm
- Cooling
- Heatsink required
Lore & Background
Matrox had been a significant player in high-end 2D graphics acceleration, with cards like Millennium and Mystique excelling in MS-DOS as well. In 1994 they released the Impression Plus, one of the first 3D accelerator boards, but it lacked texture mapping and targeted CAD. Seeing growing interest in 3D, Matrox produced the Mystique in 1997, which still lacked bilinear filtering and alpha blending. In early 1998, Matrox teamed with PowerVR to produce the m3D board using the PowerVR PCX2 chipset as a stop-gap until the G200 project was ready.
The G200 chip was used on several boards, most notably the Millennium G200 (with SGRAM and a faster RAMDAC) and Mystique G200 (with slower SDRAM and a TV-out port). Most boards shipped with 8 MB RAM, expandable to 16 MB via an add-on module. The chip featured a 128-bit core with dual 64-bit unidirectional buses (DualBus) to reduce latencies. Its 3D pipeline had a single pixel pipeline with one texture management unit, and a RISC processor called the WARP core implemented the triangle setup engine in microcode. The G200 supported trilinear mip-map filtering and anti-aliasing, though the latter was rarely used.
Reader's Guide
The G200 was Matrox's first fully AGP-compliant graphics processor, supporting Direct Memory Execute (DIME) to use system RAM as texture storage when local RAM was insufficient. In 2D, it delivered excellent speed and Matrox's renowned analog signal quality, besting the older Millennium II in almost every area except extremely high resolutions. In 3D, it scored similar to but generally behind a single Voodoo2 in Direct3D, and was slower than the NVIDIA Riva TNT and S3 Savage 3D, but remained competitive. Its 3D image quality was considered among the best due to 32-bit color depth support, assuming driver bugs were not an issue.
The G200's biggest problem was OpenGL support; for most of its life it relied on a slow OpenGL-to-Direct3D wrapper driver, which hurt performance in games like Quake II and caused controversy over delays. Early Direct3D drivers also had bugs, such as texture distortions in Unreal and mip-map flickering, which were resolved as drivers matured. The G200 series, especially the G200e, later became a popular choice for server motherboard manufacturers due to its robustness and low power consumption, often embedded in a single chip from ServerEngines and used in out-of-band management platforms.
Did You Know?
- The G200 was Matrox's first graphics processor to require added cooling in the form of a heatsink.
- It supported full 32-bit color depth rendering, which Matrox called Vibrant Color Quality (VCQ).
- The G200's 3D pipeline contained a RISC processor called the WARP core that implemented the triangle setup engine in microcode.
- A later revision, the G200A, used a 250 nm manufacturing process and did not need a heatsink at the original clock speed.
More in Graphics Hardware 1-24
Spotted an error? Know more?
Reader corrections go straight into our review queue. Suggest an edit · How this site is sourced