HP Labs find memristors can compute (w/ Video)

Apr 09, 2010 by Lin Edwards report
An image of a circuit with 17 memristors captured by an atomic force microscope. Each memristor is composed of two layers of titanium dioxide connected by wire. As electrical current is applied to one layer, the small signal resistance of the other layer is changed, which may in turn be used as a method to register data.

(PhysOrg.com) -- Researchers at HP Labs, the central research arm of HP, have discovered that a resistor with memory, a “memristor” can also perform logic operations. This means chips storing data may also be able to carry out computations without the need for a central processing unit (CPU). The discovery could mean computers will be able to become more compact and efficient than imagined previously.

HP Labs first demonstrated the in 2008 (see the PhysOrg article), but its theoretical existence was postulated in 1971 by Professor Leon Chua of the University of California at Berkeley. He named them memristors since they combine the of a memory element and a resistor. They are able to retain a memory of the amount of charge that flowed through them after the power is turned off, and the resistance depends on the voltage history. According to HP Labs these tiny devices are the fourth basic building blocks of electronics, the others being resistors, capacitors and inductors.

This video is not supported by your browser at this time.

Now a team of researchers from the HP Information and Quantum Systems Laboratory at Palo Alto, California, have created architectures for using memristors, including one in which a stack of memristors are layered in a single chip.

The team, led by the director of the laboratory, R. Stanley Williams, believe devices using the element could be developed for commercial use within a few years. These could allow supercomputers to work dramatically faster than today, and because memristors retain their memory, computers based on them could be turned off and on like a light switch.

In a news release HP Labs said memristors are faster than current storage devices such as flash memory, and they use less energy and can store more than double the data. They could also be used in handheld devices with ten times more embedded memory than today’s devices. HP Lab predicts memristor-based processors could eventually replace silicon in uses such as e-reader display screens.

Memristors are also unaffected by radiation, which can affect transistor-based technologies, and this means they could enable the development of smaller and more powerful devices, since the use of transistors is limited by Moore’s Law, which says the number of transistors that can be packed into a chip for a fixed cost has doubled every two years. Transistors on the most advanced chips have feature sizes as low as 22 nanometers, but there is a limit to how small they can become.

Williams said that as “our brains are made of memristors,” the discovery could lead to computers that work more like human brains.

The paper was published in Nature last week.

This video is not supported by your browser at this time.


Explore further: Study reveals new characteristics of complex oxide surfaces

More information:
* Nature paper: www.nature.com/nature/journal/… ull/nature08940.html
* Memristor FAQ

Related Stories

Memristor chip could lead to faster, cheaper computers

Mar 17, 2009

(PhysOrg.com) -- The memristor is a computer component that offers both memory and logic functions in one simple package. It has the potential to transform the semiconductor industry, enabling smaller, faster, cheaper chips ...

HP Licenses Technology to Create Nanoscale Electronic Devices

May 02, 2007

HP today announced that it is beginning to reap returns from its 10-year investment in nanoscale electronics with the licensing of technology that could enable the fabrication of semiconductor chips significantly more powerful ...

Recommended for you

A new way to make microstructured surfaces

1 hour ago

A team of researchers has created a new way of manufacturing microstructured surfaces that have novel three-dimensional textures. These surfaces, made by self-assembly of carbon nanotubes, could exhibit a ...

Tough foam from tiny sheets

20 hours ago

Tough, ultralight foam of atom-thick sheets can be made to any size and shape through a chemical process invented at Rice University.

Graphene surfaces on photonic racetracks

Jul 28, 2014

In an article published in Optics Express, scientists from The University of Manchester describe how graphene can be wrapped around a silicon wire, or waveguide, and modify the transmission of light through it.

User comments : 13

Adjust slider to filter visible comments by rank

Display comments: newest first

DGBEACH
3 / 5 (1) Apr 09, 2010
One question they didn't answer is how long it takes for a memristor to remember (ie. how long must the charge be present). The first time I heard about this I thought it was just another novelty-item...something discovered by accident in some R&D lab, having no real value...but now I'm thinking that if this device can at least match the speed of today's flash-ROMS, that they have just revolutionized computing!
gopher65
2.5 / 5 (2) Apr 09, 2010
DGBEACH: It'll most likely be decades before these things are sophisticated enough to change computing in a significant fashion. I'm sure there will be limited applications before then, but nothing truly revolutionary.
Morpheusjrjrjr
not rated yet Apr 09, 2010
I think 10 years will be enough to implement this technology in semiconductor manufacturing, if it will be so useful and effective.
Tim
5 / 5 (1) Apr 09, 2010
I think there could be significant applications for this technology within a relatively short period of time. Darpa is pretty keen on this tech and I have a feeling some of their funding is behind several of the recent publications about memristor technology.

"SyNAPSE is a project supported by the Defense Advanced Research Projects Agency (DARPA). DARPA has awarded funds to three prime contractors: HP, HRL and IBM. ‘SyNAPSE’ is a backronym and stands for Systems of Neuromorphic Adaptive Plastic Scalable Electronics. The stated purpose is to 'investigate innovative approaches that enable revolutionary advances in neuromorphic electronic devices that are scalable to biological levels.'"

Brace yourselves for the intelligence explosion!
DGBEACH
not rated yet Apr 09, 2010
have discovered that a resistor with memory, a “memristor” can also perform logic operations

But how could this thing perform a logical operation on its own?
sender
not rated yet Apr 09, 2010
crossbar structures in optical photoswitches would do a lot more as well
Bloodoflamb
not rated yet Apr 09, 2010
have discovered that a resistor with memory, a "memristor" can also perform logic operations

But how could this thing perform a logical operation on its own?

http://www.nature...940.html

You know... The link at the end of the article might help...
purringrumba
1 / 5 (1) Apr 09, 2010
Bringing a new technology into computation market is not an easy thing. There are many known microelectronic devices that can compute, communicate, and store information. But the mainstream technologies have been so refined and perfected that making investment to develop a new technology to compete with them is very risky and expensive: typically $400M-$500M investment (not to mention cost of re-engineering other systems and software that will have to work with this new technology).

Unless someone (or some company) figures out 1) how to champion this technology and make that kind of investment without going broke or 2) find a small niche application that is uniquely enabled by this technology, it this will likely remain an academic interest.

I don't have anything against this technology per se. I'm just stating how market is these days.
Bob_Kob
1 / 5 (1) Apr 09, 2010
Well if it is faster and cooler than current cpus the id imagine the market would be huge to take it up. It just depends on how revolutionary it is. If its only a little better then it probably wont be very successful.
sender
5 / 5 (1) Apr 10, 2010
http://einstein.s..._101.pdf

22 years from theory to practice, check page 9 for reference ;)
abouttime
not rated yet Apr 10, 2010
If DARPA is really interested they may provide a substantial amount of funding, thereby kickstarting it just as defense funding did for the internet.
Husky
not rated yet Apr 10, 2010
this has great potential for massive parallel asynchronous computing, without the need and the overhead of central cpu clockcycle and threadsyncing, even if the clockcycles are low, every memristor logic gate, has its data cache inside it, sure beats competing fetching calls from DRAMM????
canuckit
not rated yet Apr 10, 2010
This will render FeRAM obsolete.