On Lloyd Thorndyke's first day as president of a new company two years ago, he had 128 employees and backing of $40 million. ETA Systems Inc. clearly was not a garage start-up. Control Data Corp. (CDC), a $5-billion company, had hatched it almost fully grown, after deciding that creating a small company would enable it to successfully develop the next generation of supercomputers.

Pushing new ventures away from the corporate fold is a growing trend among larger companies that are trying various decentralization strategies to improve their flexibility, responsiveness, and creativity. But the arrangement cannot be effective if the parent company is essentially bundling up its own problems and sending them out the door. That may, in fact, be the case with ETA.

In launching ETA, Control Data was undoubtedly hoping to duplicate the success that IBM Corp. had in developing personal computers through an autonomous division. CDC even went a step further, with ETA employees owning 10% of the start-up.But while IBM was seeking to enter a new market with just a few strong competitors, CDC was reaching for ETA as if it were a life raft in choppy waters.

When it decided to start the new company, CDC had already fallen far behind the competition. Cray 'Research Inc., CDC's arch rival, had sold close to 100 of the 130 supercomputers that had been installed worldwide. The Japanese powerhouse Fujitsu Ltd. and many precocious start-ups have also become strong rivals.

At conception, parent and offspring were destined to miss the start of a major expansion in the market. Customers will buy about 50 supercomputers this year and an estimated 65 next year, almost doubling the total installed base. With only an aging line to offer, ETA and CDC will limp along until their new model arrives in 1986.

By then, there may be other products with more technological razzle-dazzle than what ETA has in the works. Like sports-car buffs, supercomputer buyers hunger for raw speed. "The question is always, 'How can we get more performance?" says Thorndyke. The answer may be gallium arsenide circuits, which are five times faster than presently used silicon chips. ETA's model won't use the new technology, but Cray Research is expected to unveil a gallium arsenide computer in 1987.

Given these burdens, perhaps it is no surprise that Thorndyke has been returning with empty pockets from his search for new capital. CDC anticipates outside sources will come up with about $60 million, reducing the parent company's ownership in ETA from a majority to a minority position in a few years. So far, there are no takers. "I think it will be tough sledding," says William Easterbrook, a Kidder Peabody & Co. vice-president. "They seem to be late in the game."

No small-company magic could make up for that. An entrepreneurial setup can't automatically overcome its inherited weaknesses.