Try the new FT.com

June 21, 2006 11:41 am

In the lab: The quest for universal memory

  • Share
  • Print
  • Clip
  • Gift Article
  • Comments

Why does a computer take longer than a minute to start up, and why is a digital camera so slow at saving pictures?

It's partly because the memory chips inside them are not as good as they could be. That is why the electronics industry is hard at work on the next generation of memory technologies, in a quest for a new chip known as the “universal memory”.

There are two kinds of memory widely used in consumer electronics today.

D-Ram – dynamic random access memory, the kind usually found in desktop computers – is fast and cheap, and can be rewritten many billions of times. But any information stored in it is lost as soon as the power is switched off – one of the reasons why a PC has to spend a minute rebooting when it is switched back on.

Flash memory, on the other hand, is “non-volatile”, meaning it can store data when the power is off. This makes it suitable for devices with limited (or non-existent) power supplies, such as mobile phones, cameras, smart cards, and music players such as the iPod Nano.

On every other measure, though, it is inferior to D-Ram – it is more expensive and records data more slowly, hence the delay when taking digital photographs. And after a million or so rewrites it wears out.

The search is on for a chip that will do both jobs, hence the notion of universal memory. It must be cheap enough to compete with D-Ram, and withstand being written to many thousands of times a second, but it should still store its data when the power is off.

It is a market that could be worth $75bn by 2019, according to research company iSuppli.

Several new technologies are being developed to meet this need. Probably the best-established answer thus far is ferroelectric Ram, or F-Ram. This stores information by changing the orientation of the magnetic fields of tiny portions of a material called lead zirconate titanate.

Colorado-based Ramtron has already sold hundreds of millions of these devices, which may sound impressive but is tiny compared to D-Ram shipments.

The car industry is a big customer, using F-Ram in advanced airbags, as they need a non-volatile memory that can be rewritten millions of times, to keep a constant log of a car’s speed.

Texas Instruments has also produced samples of its first F-Ram products, based on Ramtron technology. It sees F-Ram as an important technology for building a memory module into chips that do other things, such as microcontrollers and processors.

It is much cheaper to add F-Ram to a processor chip than D-Ram, says Ted Moise, F-Ram program manager at Texas Instruments, but it is just as fast. Also, the memory does not need a constant power supply, so it consumes much less power.

An even newer technology uses extremely small carbon filaments, just a few molecules thick, known as nanotubes. The filaments are suspended above a metal plate, like a diving board above a pool. When it is in the “off” state, it stands away from the pool. To switch it on, an electric current pulls it close to the plate.

At this range, nano-scale electromagnetic forces, known as van der Waal’s forces, can keep it in place indefinitely – thus making the Ram able to remember data when the power is switched off.

It sounds complex, but Nantero says it is relatively cheap to make. The nanotubes can be painted on to normal silicon in a very thin coat, then burned off using standard semiconductor lithography techniques, says Greg Schmergel, Nantero’s chief executive officer.

Nantero’s prototypes have proved as fast as existing D-Ram technologies, and potentially faster. “The question is, how to test it,” says Mr Schmergel. “We’ve tested it up to three nanoseconds, and we’re working out how to test for faster speeds.”

Nantero hopes to have the first devices on sale within a couple of years.

Another strong contender is “phase change memory”. This incorporates a thin layer of a material called chalcogenide, used in rewritable CDs, on to a silicon wafer. In response to an electrical current, small dots of this material can switch between two states, amorphous and crystalline, to represent the 0s and 1s of computer data.

A California-based company called ECD Ovonics developed the technology, and is now working with an impressive roster of semiconductor industry players, including Samsung, Intel, IBM, Siemens, and STMicroelectronics to make commercial versions.

There are also many others at various stages of development. Freescale, the semiconductor company spun out of Motorola, is in the first stages of commerialising a technology called magnetic Ram, which takes magnetic materials similar to those found in hard disks and tapes, and integrates them into silicon chips.

It is also working on nanocrystalline Ram, which remembers data by storing electrical charges on tiny silicon crystals.

New memory technologies are already taking off in niche applications, and it is likely that they will continue to expand those niches and find new ones.

But the mainstream market for electronic components is extremely price-sensitive. The new memory technologies will not begin to displace old fashioned D-Ram and Flash until they can compete on price, and it will take billions of dollars of research and development spending to get to that point.

Of all the rival technologies, phase change memory seems to have the best chance of prospering, according to Mark DeVoss, senior analyst at iSuppli.

“You have to look at the number of partners to see who has the best chance,” he says. “Phase change memory has a lot of big companies behind it, with a lot of money to spend on research and development.”

This vast effort must eventually succeed, as the industry demands ever cheaper, faster and more flexible components and soon existing Flash and D-Ram technologies will not be able to keep up.

“Something has to happen,” says Mr DeVoss. “They are going to need some new technology soon.”

Copyright The Financial Times Limited 2017. You may share using our article tools.
Please don't cut articles from FT.com and redistribute by email or post to the web.

  • Share
  • Print
  • Clip
  • Gift Article
  • Comments
SHARE THIS QUOTE