Unveiled at Siggraph 1996, Talisman was a Microsoft initiative to improve the quality, performance and integration of audio and video media technologies on the PC. Its eventual aim was a single PCI expansion board which integrated audio, video, 2D and 3D graphics and MPEG-2 decoding. Rather than being a refinement of existing technologies, the graphics element of Talisman was an attempt to define a new, fundamentally different architecture, which solved the problems of bandwidth limitations, low frame rates and high latency/low interaction which had previously held PC-based 3D graphics back.

Microsoft described it as the replacement of image synthesis with image processing, based on specialist DSPs. Instead of the usual graphics pipeline, with the images created from primitives, rasterised and sent to the frame buffer as a whole, the Talisman architecture created a series of separate image layers that could be rendered and manipulated independently. Instead of updating the entire image for each frame, each layer was capable of being updated as required, based on priorities set up by software. For example, hazy objects in the background of a 3D scene could be updated less regularly and less accurately than clearly visible objects in the foreground. Since not all image layers needed to be updated at any one time, a significant saving on processing time and bandwidth could be achieved. The rendering of objects into layers also allowed 3D transformations to be substituted for by 2D image operations, as long as there was not enough distortion to warrant a complete re-rendering.

Talisman also used a process called chunking, involving the breaking down of each image layer into 32×32 pixel regions called chunks. All the geometry in one chunk was rendered before proceeding onto the next. As the Z-buffer only had to deal with one chunk at a time, it needed only to be as large as a single chunk, and as a result it was small enough to be implemented directly on the chip. Chunking enabled a block-oriented image compression: once rendered and anti-aliased to remove jagged artefacts or pixelisation, each chunk could be compressed, saving further on resources.

Initially, Microsoft was supported by Samsung, Fujitsu and Cirrus Logic, all of whom pledged silicon to implement the technology. However, it was not long before Talisman found itself was overtaken by events. Intel boosted CPU and RAM speeds and announced the AGP bus, the chip vendors that had previously supported the project dropped out one by one, and the 3D graphics companies seized the initiative and started to do what they do best.

The result was that rendering philosophy was in fact defined by the likes of 3Dfx, and not by Microsoft’s somewhat myopic vision of the future. In the end the Talisman technology was licensed by 3D semiconductor company Trident Microsystems, and Microsoft subsequently migrated some of its simpler features to its DirectX technology.