Tech Industry

Another start-up to try Intel clone chip

It's so far proved to be the impossible dream in the chip market: to make money selling a chip based on the same architecture as Intel's chips. But another start-up is going to try.

It's so far proved to be the impossible dream in the semiconductor market: to make money selling a chip based on the same architecture as Intel's chips. But another start-up is going to try.

At the Microprocessor Forum in San Jose, Calif., this week, start-up MemoryLogix will unveil its blueprints for a chip designed for cell phones and handhelds and based on the x86 architecture, which underlies the majority of Intel processors.

MemoryLogix's designs describe a chip that runs at only 400MHz but is energy efficient and incredibly small, said Kevin Krewell, executive editor of industry newsletter the Microprocessor Report, which sponsors the chip industry convention.

"It is very competitive with the ARM stuff," Krewell said, referring to the main chips found in cell phones today, made by British chip designer ARM Holdings. "You are not going to put it into a PC, but it is certainly an acceptable performance level for an embedded application like a handheld."

Not much is known about MemoryLogix at the moment, but more details will emerge Wednesday when company president Peter Song, a former contributor to the Microprocessor Report, makes his presentation at the conference. The forum is one of the major events on the semiconductor social calendar, with many companies using it to unveil new products.

According to the program notes for the conference, MemoryLogix's new x86 core "supports advanced micro-architectural features typically found on Pentium-class processors, yet is no larger than some low-cost embedded CPU cores." MemoryLogix could not be reached for comment, and the company's Web site has yet to be unveiled.

Theoretically, it's a lucrative opportunity. The x86 architecture has been the basis of the vast majority of chips produced by Intel since 1971, when the company released the first-ever microprocessor, the 4004. Most desktops, notebooks and servers contain x86-based chips, and many products and some portable products, such as Research In Motion's BlackBerry wireless device, do too.

Analysts and semiconductor executives have also predicted that x86-based chips could begin to creep into the portable market. Chips based on designs from ARM currently dominate this market--Intel itself sells one--but they can't run standard PC software.

With an x86-based chip in a handheld, developers wouldn't have to modify their desktop applications to run on ARM chips (and chipmakers presumably wouldn't have to pay royalties to ARM). OQO plans to bring out a Palm-sized PC later this year that reads x86 software.

Intel, though, has managed to knock out most of its would-be competitors, or at least confine them to the fringes of the market. Historically, Intel does not license the x86 architecture. A few companies have licenses, but they are large companies, such as IBM, and the licenses were granted in massive cross-licensing deals.

Intel has filed lawsuits against other companies selling x86 chips, such as Via Technologies. And because PC makers that use chips from these companies can also be sued for patent infringement, most won't even consider using the chips until the legal dust settles.

More importantly, Intel's massive engineering and manufacturing resources give it the ability to increase production volume, cut prices and/or boost the performance of its chips almost at will to head off emerging competitors.

One example of this sort of approach involves Santa Clara, Calif.-based chipmaker Transmeta. Intel has refrained from suing Transmeta, which makes an energy-efficient chip that can run the same software as Intel processors. Instead, Intel simply came out with its own line of energy-efficient Pentium IIIs. When Transmeta faltered, customers adopted Intel's chip.

Other companies that have tried, without much success, to market an x86-based chip include National Semiconductor, Rise Technology, IBM and IDT. Advanced Micro Devices is the only company to achieve significant market share, but it has had only two profitable years since 1995 and will likely lose money again this year.

MemoryLogix's chief potential advantage lies in the size of its chip. The core of the microprocessor takes up only 3 square millimeters, said the Microprocessor Report's Krewell. Adding a memory controller and other elements increases the surface area, but the chip will still be small. Small chips cost less and generally consume less energy, critical factors in the handheld market.

MemoryLogix will not likely make the chip, Krewell added. Instead, it will license its designs to larger semiconductor companies.

Intel declined to comment.