Like a Don Quixote of computerdom, Gene Amdahl has tried to topple International Business Machines Corp. as the world's preeminent manufacturer of mainframe computers.

Twice, Amdahl, a former IBM scientist widely regarded as a brilliant computer architect and entrepreneur, has failed.

In 1970 he founded Amdahl Corp., which produced a family of high-performance, high-capacity, general-purpose business computers that competed head-on against IBM. It constantly required cash, and eventually he sold a major stake in the company to the Japanese computer giant, Fujitsu Ltd. Amdahl left in 1980.

Then he started Trilogy Ltd., an ambitious, expensive effort to produce an IBM-compatible supercomputer on a 2 1/2-inch silicon wafer. That didn't work either.

Third time lucky?

No, said Amdahl: "We have to learn to scale down expectations."

After the technological failure of Trilogy Ltd., Amdahl, 62, has "elected to find a computer company that was ready for a larger marketing effort."

But rather than compete directly with IBM's strength, Amdahl has chosen to apply his talents and capital to reenter the computer market at a more oblique angle. This time, Amdahl is gunning for the high end of the minicomputer market. Minicomputers occupy a niche halfway between mainframes and personal computers. They are a multibillion-dollar market, but one that now is dominated by companies such as Digital Equipment Corp. and Data General Corp., not IBM.

His latest technical edge? Parallel processing: computer architecture that Amdahl began exploring back in the 1960s when he was still at IBM.

Earlier this summer, Trilogy, flush with roughly $50 million in cash from coporate investors and venture capitalists despite its failures, merged with ELXSI, a San Jose company that had been manufacturing parallel-processing computers for nearly two years.

ELXSI's machine was "architecturally superior," said Amdahl, who paid about $50 million in stock for the $20 million computer firm. "It had the fewest bottlenecks and was the least dependent on high technology for performance -- which means it has the most promise."

That's important, said Amdahl, because it means ELXSI can exploit some of the technology that Trilogy devised in its futile effort to develop its own computer.

For example, Amdahl said that Trilogy created new ways to power and cool densely packed computer chips. Moreover, Trilogy asserts that it has figured out better ways to interconnect chips so that they can communicate with each other at very high speeds -- a vital function for any parallel-processing machine.

Trilogy also claims to have developed high-speed memory chips, computer memory devices capable of retrieving data in five billionths of a second -- far faster than current memory chips.

While those advanced technological features may be added to future generations of machines, the reason for the merger, ELXSI chief Joseph Rizzi conceded, was "the money," which the company needs to fund growth.

ELXSI is "striving to be profitable next year," Rizzi said, and developing and marketing a next-generation computer system is "extremely capital-intensive."

Like other "superminicomputer" companies, ELXSI is trying to fill what it sees as a gap in the computer marketplace. "We can bridge from the Digital Equipment Corp. world all the way to the Cray supercomputer world," Amdahl said.

ELXSI believes there is a price performance gap between the high-speed and expensive minicomputers offered by traditional minicomputer companies such as Digital and the ultra-high-speed supercomputers sold by Cray.

ELXSI's 6400 parallel processor offers a price/performance wedge that the company claims makes it compatible with most minicomputers, but at speeds close to supercomputer performance. Rizzi said that more than 70 organizations, including most of the aerospace companies, have bought EXLSI machines to do computer-aided design work.

Still, it is far too early to say whether ELXSI can be a survivor in the crush to provide parallel-processing power to customers, even with Amdahl's expertise and Trilogy's money.

"Opening up the market is taking longer than we thought," said Rizzi, who adds that it's nice to have a world-renowned genius associated with the company.

"Everybody wants to meet him," Rizzi said. "He's a great marketing tool."

But Rizzi and Amdahl expect that IBM will enter their market if it proves to be as lucrative as they envision.

"I don't think IBM will be happy until it owns the world," said Amdahl, who concedes that, no matter which way one turns in the industy, the world's largest computer company materializes.