Mainframe and minicomputer companies have come to realize the importance of a distributed-workstation environment by watching Sun grow to over $2 billion in 1990 and take large fractions of their markets. By 1990, DEC and HP share less than half the market, and as the year has progressed, their share has dwindled even further. HP's divisional structure is organized to build and sell competitive 68000-based workstations. In 1985, HP began using its own RISC chips, and it bought Apollo in order to have a larger installed base of customers. DEC took longer to enter the market, given its need to have VAX-on-a-chip processors as its workstation engine. In 1989, DEC introduced a high-performance UNIX workstation using the MIPS RISC chip as its "Sun killer," followed by a faster version in the summer of 1990. Sun responded with its SPARCstation, and Data General entered the fray with a high-performance Motorola 88000-based workstation. In 1990, Evans and Sutherland made its first entry into the workstation market with a high-performance 3-D workstation based on the MIPS microprocessor to challenge its progeny, SGI.

In the late 1980s, IBM introduced several technical Unix workstations in the RT series (IBM's RISC architecture) and one based on the 68000. By 1990, none of IBM's products was powerful enough to appeal broadly to the technical and software-development community. Andy Heller, former president of IBM's Workstation Division, offered the following observation about the effect of having a long gestation period before a product finally becomes available: "Technology is like fish-the longer they stay on the shelf, the less desirable they become." In February 1990, IBM introduced a range of UNIX-based workstations and servers that is likely to retain a product edge for one to two years. These products are based on IBM's next-generation, superscalar RISC technology, which provides substantially higher performance for technical applications than ordinary RISC processors.

Given the push for performance as the differentiator, all companies have chosen different strategies for achieving supremacy. The multiprocessor approach introduced by Apollo, Stardent, and Silicon Graphics allows an arbitrary amount of power to be placed in a given computer. With several firms-such as DEC, Data General, and various Intel 486- based workstation companies-introducing such products in 1990, the 1990s will truly see the advent of parallel processing.

Some further observations on the performance wars:

These efforts rarely affect products and are often the basis of a start-up such as Apollo.