Feature connector

From Wikipedia, the free encyclopedia

The Feature connector was an internal connector found mostly in some older VESA Local Bus, ISA and PCI graphics cards, but also on some early AGP ones.

It was intended for use by devices which needed to exchange large amounts of data with the graphics card without hogging a computer system's CPU or data bus, such as TV tuner cards, video capture cards, MPEG video decoders, first generation 3D graphic accelerator cards and the such.

Several standard existed for feature connectors, depending on the bus and graphics card type. Most of them were simply an 8, 16 or 32-bit wide internal connector, transferring data from and to the graphics card to another device, bypassing the system's CPU and memory completely.

Their speeds often far exceeded the speed of normal ISA or even early PCI buses, (e.g. 40 MB/s for a standard ISA-based SVGA, up to 150 MB/s for a PCI or VESA-based one, while the ISA and VESA bus had much lower or barely as much bandwidth, with such figures being far beyond the capabilities of e.g. a 386, 486 or barely handled by an early Pentium.

Depending on the implementation, it could be uni or bi-directional, and carry analog color information as well as data. Unlike analog overlay devices however, a feature connector carried mainly data and essentially allowed an expansion card to access the graphics card Video RAM directly, although directing this data stream to the system's CPU and RAM was not always possible, limiting its usefulness mainly for displaying purposes.

Although its use rapidly declined after the introduction of the faster AGP internal bus, it was, at its time, the only feasible way to connect certain types of graphics-intensive devices to an average computing system without exceeding the available CPU power and memory bandwidth, and without the disadvantages and limitations of a purely analog overlay.

The idea of accessing a video card's memory directly has recently resurfaced with the introduction of the Scalable Link Interface (SLI) technology by NVidia, and less recently by 3dfx's equivalent Scan-Line Interleave, although those technologies are mostly aimed at connecting two equally powered and complete graphic cards in order to produce a single, increased performance visual output, and not e.g. directly interfacing TV tuner cards.