Approaching the desktop supercomputer

The history of desktop supercomputing, the current status of the technology, and the problems still remaining to be solved are examined. Desktop computers fall into two major categories, workstations and PCs. Workstations essentially can be viewed as large-scale computing systems cut down in size and capability to yield a useful desktop package. They invariably run a variant of the Unix operating system. PCs have traditionally been distinguished from workstations by two factors: a simplified operating environment and easy extensibility. RAM, disks, and other peripherals can be easily and inexpensively added by a user with virtually no technical expertise. One of the earliest approaches to increase the computing power of a PC was to use array processors (APs) to perform that portion of the calculations that was enhanced by the vector concurrency or the fast static RAM available on the APs. The early PC AP systems were derived from the mainframe and minicomputer AP technologies. The scalar computing performance of a desktop PC equipped with a computing coprocessor system is essentially the same as that of typical mainframes and approaches that of supercomputers. Computing coprocessors offer the highest levels of performance with simplicity of operation, bringing supercomputer power to the desktop without adding complexity to the well-accepted PC software environments.<>