Introduction to G400

Last update: 1999-06-24

This is my informal introduction to G400 technology. It may be useful for people considering buying the G400-based card, as it explains the most interesting, fresh features of the chip.

Memory interface

G400 works with up to 32 MBytes of SDRAM or SGRAM memory. The interface is 128 bits wide, and is clocked at 125..200 MHz. This effectively means that memory bandwidth no longer imposes any serious limitation on screen resolution and refresh rates. Theoretically G400 can run any resolution 32 bpp display at 80..85 Hz screen refresh.
The high speed of memory interface has one side effect - the video memory is not ugradeable, since no memory extensionmodule could easily comply with the physical transmission requirements.

DualHead

G400 chip is a heart of single video subsystem with two independent outputs. There is only one host interface and one drawing engine, but there are two data-hungry sequencers willing to suck pixels from video memory and send them to the screens. This has an advantage of two screens connected to single computer, but it also has a disadvantage of reducing the throughput of memory as viewed by any of the logical components. Using two screens reduces the limits of the primary display (screen refresh at high resolutions) and also reduces the drawing speed.

The sequencers

The two sequencers are different: first is an enhanced VGA-like sequencer, similar to that found in previous chips, including G200. The second one is much simpler (and easier to program), and it supports only the features that are really needed in contemporary video subsystems.
CRT2 (the symbol of second sequencer) works with 16-bit and 32-bit pixels in direct color, RGB mode. It can also display the contents of memory interpreted in a way used for video processing - YCrCb. In "video" modes the three values may be stored in separate planes. Note that CRT2 cannot work with palettized colors (no 256 color mode).
The second sequencer is intended for TVoutput. It is also suitable for Flat Panel displays.

The RAMDAC

Although the G400 chip contains two sequencers, it contains only one RAMDAC. Usually the primary sequencer sends the data to RAMDAC, but one can program the G400 in such way that CRT2 is connected to internal RAMDAC.
This RAMDAC is the direct descendant of Hurricane RAMDAC introduced in old good Mystique chip. Besides higher speed, its characteristics and features remain almost unchanged since MGA-1064SG chip.
So what's happening to the secondary pixel stream? It is transferred throug 12-bit wide digital interface "somewhere outside" of the G400 chip - to digitally controlled flat display, or to some for of external RAMDAC. Martrox has a ready solution - the next incarnation of MAVEN chip, known from Mystique G200, found on all the DualHead boards.

MAVEN

MAVEN stands for MAtrox Video ENcoder, and is a name of family of chips used on Matrox cards. The MAVEN chip used on G400 cards accepts 24 bit video data sent in 12-bit portions, stores it, performs advanced digital filtering and outputs the data in analog form, either as RGB signals or as separate or composite video signals in PAL or NTSC standard.
The digital interface between G400 and MAVEN works at up to 270 MHz. This results in maximum pixel clock limit for the "second head" display of 135 MHz, which is still quite a lot - up to 1152x864 at 85 Hz or 1280x1024 at 75 Hz.