The architecture of graphics systems is at a major crossroad. The large market for games and entertainment applications has resulted in major investments in graphics chip technologies. The result is that state-of-the-art graphics chips (e.g. the NVIDIA GeForce2 and the SONY PlayStation 2000 EE and GS chips) have similar complexity and greater computational density than current microprocessors. The challenge now for the scientific community is to build high-performance scalable graphics systems out of inexpensive, high performance components. Moreover, since graphics capabilities are now widely available, it is time also to rethink how graphics resources are composed and partitioned across a variety of networks (SAN, LAN, WAN).
In this project we are investigating technologies for building scalable graphics systems using commodity components. Specifically, we want to develop efficient systems for driving new display configurations such as tiled arrays of projectors, for generating and receiving remote graphics streams, and for partitioning and reconfiguring heterogeneous graphics and display resources. The resulting set of capabilities we call multi-graphics.
As part of this project, we are building a visualization cluster consisting of 32 PCs with commodity graphics cards.