TOKYO — After a year of uncertainty, the embedded StrongARM processor has found a home with its new parent, Intel Corp., which is scheduled to have its first home-grown version of the processor ready by the second half of 1999.

The company has come a long way since it announced it would buy Digital Equipment Corp.'s fabrication facility in Hudson, Mass. as part of its antitrust settlement with Digital, a deal that gave it rights to StrongARM.

Some analysts remain skeptical.

"Intel has never before designed a StrongARM processor," said Jim Turley, a senior analyst at Microprocessor Report. "And there is some concern that the team may have trouble duplicating the impressive results of the original StrongARM team, which has since disbanded."

Despite these concerns, Intel said it has regrouped. It has been hiring engineers at its embedded division in Chandler, Ariz., and has started determining the specifics of the next StrongARM. "We have already made the team for the next-generation core. We have tens of people in the design team and we now have a support team," said Itaru Sonomoto, senior engineer for Intel K.K.'s StrongARM team and formerly with Digital. He said Intel is also in the process of adding more people to its team in Japan, which is considered one of the key markets for StrongARM.

At the same time, the company has set a course of continuity with Digital's StrongARM strategy of integration. Intel has traditionally spurned adding peripheral functions around a processor core because it dilutes the value of the processor, Turley said. But in the case of the StrongARM, Intel appears to have embraced more integration just as Digital had for the current generation of Strong ARM 110, 1100 and 1500 processors. That strategy seems to have paid off. Hewlett-Packard Co. is using Intel's 190-MHz 1100 device as the processor for its newest Jordana 820, a Windows CE-based machine.

"In the future we will have the SA-2 with a cache memory management unit and maybe several peripherals and co-processors," Sonomoto said. "It's the same strategy as the SA-1 processor."

Still, Intel has some reservations about how far it can go with integration. Some of StrongARM's biggest competitors, most notably Hitachi Ltd.'s latest SH processors, come equipped with flash memory on the same die. But Intel has no immediate plans to integrate flash on the die.

Said Takahiro Katoh, a marketing manager for Intel K.K, "We're looking to offer a standard platform, which can run any of the available software."

Turley said he's not surprised that Intel is hesitant to integrate flash. "StrongARM is valuable only when it runs really, really fast. It runs fast because Digital made it — and still does — on its advanced semiconductor-process technology in Massachusetts. By its nature, SA needs to be built on a leading-edge process, and polluting that process with flash memory would slow the chip down considerably. Mixed memory and logic runs slower than pure logic processes."

Yet it's not a simple issue for Intel. While the company may gain an advantage in its process technology, some have recognized that it may lose its edge in power consumption in mobile applications.

"Many customers have been saying they want [integrated] flash," Sonomoto said. "It might be useful in mobile-phone applications. If we integrate flash or mask ROM, we can save on power dissipation for mobile-phone applications. Otherwise, every time the StrongARM wakes up it will have to go out to external flash."

Low power consumption has been one of the hallmarks of StrongARM, and Intel intends to make it one of its biggest selling points. When Intel rolls out the SA-2 in late 1999 or early 2000 as it expects, the chip will have considerably higher performance than its predecessor — running faster than 300-MHz — but will not exceed the 250-milliwatt power-consumption ceiling of existing StrongARM chips, Intel officials said.

To meet those parameters, Intel is evaluating whether to use its most advanced 0.18-micron technology from the outset rather than its current high-volume 0.25-micron process. The company still hasn't made a decision on process technology, but whatever the case StrongARM will likely be one the first in line for the new process, Sonomoto said.

"Intel will not drive process technology using StrongARM. Merced is the driving IC. However, StrongARM will use state-of-the-art semiconductor-process technology," Sonomoto said.

Intel is designing the SA-2 for its own process technology rather than for current fabrication technology used at the Hudson fab where StrongARM chips are now produced. Such a move would allow Intel to ramp up the devices in high volume quickly, Katoh said.

Intel also does not want to rely on a fab that uses a process technology different from its own for future iterations of StrongARM, observers said.

Porting a fresh design to Intel's vaunted fabrication process could also give StrongARM the edge it needs to quickly scale up performance, as Intel proved it can do with its Pentium processors. That should help it compete against a raft of competitors in Japan, such as Hitachi's SH or various MIPS-based processors from NEC Corp. and Toshiba Corp.

Even so, Intel officials admit that StrongARM will have to struggle against more entrenched rival architectures. This is especially true in Japan, where many of the Jupiter-class Windows CE PCs are being designed. "StrongARM is not so popular here now," Katoh said. "Marketing has just begun, and I think it will take one to two years to make gains."

Ironically, one of Intel's most worthy opponents will likely be Advanced RISC Machines (ARM) itself. The company recently debuted its ARM10 processor core, which adds multimedia processing functions and runs at 300-MHz with 420 Dhrystone Mips. The core is expected to be produced in samples by ARM's partners by the third quarter of 1999, with full production in 2000.

Intel recognizes ARM10 as a direct competitor. But the company could gain an advantage if it manages to quickly move the SA-2 core to 0.18-micron process technology before its rivals, which is conceivable, given the company's leadership in shrinking its line widths.