A flash in the pan?

A cover story in Electronics magazine, co-authored by Intel's Gordon Moore, outlines the technology behind a revolutionary memory chip that could reduce manufacturing costs, increase storage density, require little electricity and retain data forever. Researchers now say that the new memory--called Ovonics Unified Memory--could hit the market sometime toward the end of the decade.

Which would be about 40 years after the article was published.

The history of Ovonics shines light on the potential rewards, and risks, in the crazy world of nonvolatile memory, which is memory that doesn't lose its data after the power on a host computer gets switched off.

Flash, the main form of nonvolatile memory, constitutes a mammoth business. In 2001, approximately $7.3 billion worth of flash memory--used to store data and applications in cell phones, industrial equipment, MP3 players and other consumer electronic devices--was shipped, according to Jim Handy at Semico Research. By 2007, the figure will rise to $43 billion.

Every year, the amount of flash--measured in bits shipped--doubles annually, said Handy. Put it another way: Every year, more bits of flash memory are shipped than were shipped in every previous year combined.

It should be a gold mine, right? Not always. Prices can decline almost as fast. In 2002, prices dropped by 52 percent from the year before, canceling out the potential financial gain from doubling the amount of bits shipped. Intel raised prices on flash in January to capitalize on a shortage: the company lost market share" instead.

This week, Advanced Micro Devices and Fujitsu formally combined flash operations to cut costs.

To top it off, the underlying technology for flash is starting to rub up against fundamental physical limits, prompting companies to examine alternatives. A company that develops an alternative could see dollars roll in, especially in those first few years when competitors have to learn to emulate it.

The tough part is that, for the Ovonics material to work as memory, microscopic points on a sheet of the material have to be heated to 600 degrees.

Motorola says it can extend the life of flash memory by replacing an existing layer inside the chips with nanocrystals, using a thin lattice of silicon atoms. IBM is looking at magnetic RAM (MRAM), while Texas Instruments and Ramtron say the answer is in ferroelectric RAM (FeRAM).

Unfortunately, none of the alternatives are ready for production just yet. In fact, most sound a little too Wired magazine circa 1997 to be practical.

Ovonics, which Intel is pursuing, does away with the problem of storing electrons inside a transistor, which is how current flash technology works. Instead, data is melted into a sheet of DVD-like material. (This material could also be used to displace batteries, revolutionize solar cells and improve communications.)

"A new scientific approach to materials based on disorder and local order can enable the development of new pollution-free technology which will answer society's urgent needs to reduce its dependence on uranium and fossil fuels, particularly oil," wrote inventor S.R. Ovshinsky, the inventor and founder of Energy Conversion Devices. "Science and technology which can change the world's dependence on it can create new huge industries so necessary for economic growth--$30 trillion in 30 years for new electricity alone."

The tough part is that, for the Ovonics material to work as memory, microscopic points on a sheet of the material have to be heated to 600 degrees. The overall Ovonics market is still small.

MRAM involves detecting that a spinning electron has changed direction. Besides the awesome technical challenges, imagine trying to sell this to the public. If I am talking on my

Invite Michael Kanellos into your in-box Senior department editor Michael Kanellos scrutinizes the hardware industry in a weekly column that ranges from chips to servers and other critical business systems. Enterprise Hardware every Wednesday.

cell phone and rub a balloon on my head, will my call get rerouted to Irkutsk? Do I need to buy a lead apron to use a game console?

Other companies are looking at building 3-D chips that cut costs by reducing surface area. FeRAM involves shifting an atom inside a Pervoskite crystal. In other words, it's more complex than the glitter pen or a plug-in deodorizer.

Whatever the technology, the lead scientists at rival companies are generally more than willing to poke holes in it. Then again, flash itself shouldn't be possible either. "There is no reason it should work at all," said Tom Lee, an associate professor at Stanford and co-founder of Matrix Semiconductor.

So Moore shouldn't feel too bad that his prognosis went awry. Predicting the future is tough. In the same September 28, 1970, issue of Electronics, another article underscored the difficulty manufacturers could have in bringing another alleged Next Big Thing to market.

The article: "The big gamble in home video recorders."