Philadelphia (PA) – Scientists at the University of Pennsylvania claim to have found a new non-volatile nanoscale memory technology that retrieves data 1000 times faster than flash, stores data for about 100,000 years and has the potential to scale into the terabit range.

There is no doubt that there will be a successor to flash memory in the not so distant future, as the technology is moving closer to limits of economics and scalability. The question, however, remains which memory technology will be succeeding flash – and when it will be a viable option.

Researchers at the University of Pennsylvania have added another possible approach to the growing list of possibilities. The group of Ritesh Agarwal, an assistant professor in the Department of Materials Science and Engineering, has developed a variant of phase change memory, describing a technology that is based on developed a self-assembling nanowire of germanium antimony telluride, a phase-changing material that switches between amorphous and crystalline structures, the key to read/write computer memory.

Other than current silicon manufacturing techniques that involved lithography, these nanowires, roughly 100 atoms in diameter, use self-assembly, a process by which chemical reactants crystallize at lower temperatures mediated by nanoscale metal catalysts to spontaneously form nanowires, the scientists said. The nanowires were 30-50 nanometers in diameter and 10 micrometers in length.

First test results are promising as Agarwal's group claims that data writing, erasing and retrieval processes lasted only 50 nanoseconds - which would mean these devices are around 1000 times faster than today's average flash memory devices – while consuming very little power in data encoding – about 0.7 mW per bit. And apparently, the technology could be very durable: "The device would not lose data even after approximately 100,000 years of use," the group stated in a press release.

There are a few promising flash successors out there, but it appears that most of them at least have one major drawback. Often it is a requirement of high storage density to allow a new technology to compete with the enormous capacities of flash devices today. Ritesh Agarwal told TG Daily that his phase change approach has the "potential" to scale to "terabit-level", which is what flash memory devices are slowly but surely taking aim at already today. However, Agarwal noted that the first fundamental concept of the technology has been demonstrated and the project is moving into a phase of putting all the pieces together. But he stressed his belief that this technology will be able to compete with flash in terms of density, (if and) when it becomes available.

The downside of this technology, at least for consumers, is that it is far from hitting the market. Agarwal told us that a commercial product is still eight to ten years away.

.7mW per bit? Let's see 8 bits per Byte = 5.6mW pe...

Sep 17, 2007 14:24     

.7mW per bit? Let's see 8 bits per Byte = 5.6mW per Byte. 1000 Bytes= 5.6 Watts? yikes!
MIKE report abuse vote down vote up Votes: +0

Good thing it doesn't work that way

Sep 17, 2007 14:50     

5.6 mW per 1000 Bytes IF you were Writing all 1000 Bytes Simultaneously. As it is it is still only 0.7mW for the duration of writing 1000 Bytes, which I can only imagine won't take very long if this technology perfoms the way they claim it does. You have to differentiate between instantaneous power consumption and total power consumption.
Johnboy report abuse vote down vote up Votes: +0

Re: 5.7 Watts per kB

Sep 17, 2007 14:53     

Sure it's 5.7 Watts per kB, or 6 MWatts per GB, but times 50 nanoseconds... that means to write/read/erase one gig only uses 8.3E-5 KWatt/Hours. Not so shabby. (This could be way off, but the fact that it uses 5.6 Watts isn't a big deal if it's in very very short bursts)
Kellen report abuse vote down vote up Votes: +0

feasible?

Sep 17, 2007 14:54     

I agree, 5.7 Watts per MB is really high. That's 5.8kW per GB. But my question is this... How readily available is germanium antimony telluride? Where do we get this, how is it synthesized, how much power is required in the manufacturing process? Let's face it, we can build cars out of Titanium but it's too damn expensive and there isn't enough Titanium to go around. Is this the same scenario?
cheepgeek report abuse vote down vote up Votes: +0

Wish there was more info on the power...

Sep 17, 2007 14:58     

5.7344 watts per kilobyte? It seems like a lot, I guess it comes down to how long it has to use that much power for. Maybe it saves power because it's so fast, seems like a lot of draw, though I am not familiar with storage media power requirements.
Jwolf report abuse vote down vote up Votes: +0

This seems GREAT!

Sep 17, 2007 15:31     

if they don't com up with something better or cheaper before production in about 5-8 years, As for the cars, Titanium is a poor structural metal... its way to flexible and it expands a lot with temperature.
Gabe report abuse vote down vote up Votes: +0

Can someone please give those guys these tera $...

Sep 17, 2007 15:39     

so that they can hit the market sooner? Gosh, how badly I would love to get rid of all the HDDs, DVD-Rs and Flash cards and replace them with Solid State drives that would not break mechanically and not loose data after a couple of years of usage!
paulie report abuse vote down vote up Votes: +0

how many re-write cycles?

Sep 17, 2007 17:32     

flash fails after around 100,000 right?
Ted report abuse vote down vote up Votes: +0

Great but not so great

Sep 17, 2007 21:03     

On one hand this is interesting. But the problem here, as with deploying antimony in III-V materials for solar energy is that based upon US Geological Survey estimates, the world only has about 30 years of antimony left in reserves using at the current rate (http://pubs.usgs.gov/of/2003/of03-019/). Start using this up and watch the price go up dramatically. If this is 8-10 years away, we are down to roughly 20 years worth asssuming there is not an increase in consumption. With China and India growing like they are, it becomes easy to envision a supply less than 30 years.

TheOtherBubka report abuse vote down vote up Votes: +0

Clarifying power ...

Sep 17, 2007 21:05     

Think about the power this way: writing one bit at a time, at 50ns per bit, means a write speed of 2,500,000 bytes per second, while only consuming 0.7mW the whole time. Now can you see just how good the tech is? Germanium and Antimony are fairly common, Tellurium less so - but remember that very little material is required to make nanoscale devices!
Twospoons report abuse vote down vote up Votes: +0

Another flop

Sep 17, 2007 23:28     

Some great tech comes up that requires a massive loan to pay for the electric bill. Even if it is super fast, this still does not seem to be the answer to flash problems. Still, "use self-assembly," makes it sound like it's alive or something. One step closer to creating life?
Joe report abuse vote down vote up Votes: +0

Vista uses flash memory to increase main memory for...

Sep 18, 2007 01:21     

Vista uses flash memory to increase main memory for starting applications. How does this kind of memory compare to current and maybe future RAM speedwise? There was a discussion not long ago about non volatile memory being so small and fast it could replace caches and HDDs alltogether. Is this any step in that direction?
Virtualban report abuse vote down vote up Votes: +0

Christ...

Sep 18, 2007 09:30     

If the bits are written in serial with no other control data or interuptions then: time * num bytes in 1GB * bits in a byte 0.00000005 * (1024*1024*1024) * 8 = 429.4967296 seconds or 7.158278827 minutes or 0.119304647 hours 0.119304647 hours * 0.0057 Watts = 0.000680036 Watt Hours As the maths is all multiplicative so if the bits are written as bytes (8 parallel) then it would be 8 times the watts for 1/8 of the time. Which would be 8/8 or 1 which is the same amount of overall energy used but obviously faster results

Tony report abuse vote down vote up Votes: +0

Antimony

Sep 18, 2007 09:34     

I'm not familiar with this material. Can it be recycled/recovered from obsolete/failed equipment?
kw report abuse vote down vote up Votes: +0

re: Christ

Sep 18, 2007 11:30     

and to put 0.000680036 Watt Hours into real day perspective. A unit of electricity on your bill is 1 KiloWatt Hour To reach that useage u would have to repeat the 1GB operation 1470509.328 times. In real life though the device will not just be made from the memory but other components like a controller etc. This will increase the Wattage and energy cost but you get the picture.
Tony report abuse vote down vote up Votes: +0

'self assembling'

Sep 18, 2007 11:50     

I read an article some years ago about what is called 'top down' fabrication. That is a description of how current integrated circuits and microchips are produced. 'Top down' means etched. The article suggested that a hard limit in physics has been found, in the lab, and that new and radical manufacturing techniques would need to be developed. Enter 'bottom up' concept, where we genetically engineer a microbe that cannot survive in the natural environment. In certain conditions, it will, for example, coat itself in a conductor, attach it's ends to something (completing a circuit) and promptly die. Otherwise known as 'self assembling'. The potential for such organisms to be created, and the potential they have for destroying or heavily changing living conditions here on earth worries me. I hope we are responsible with such a technology. Perhaps we have found a good use for the moon.

Scott report abuse vote down vote up Votes: +0

Sweet

Sep 18, 2007 11:52     

Love the commenting system!
brian report abuse vote down vote up Votes: +0

8-10 years ???!!!

Sep 18, 2007 15:55     

This article had me excited until the last sentance. 8-10 years in the computer/electronics industry is a lifetime. It means this tecnology is basically useless and is probably just a theoretical posibility at this time. It will be obsolete in 8-10 years.
Bob report abuse vote down vote up Votes: +0

Many other factors to address

Sep 18, 2007 20:06     

A lot of good questions and comments about the power consumption and availability of the natural resources necessary have been noted, but we should be sure to question some other factors that may or may not be issues: 1) Volatility - Although they claim 100,000 years, we can all agree it's a claim that is hard to prove. What real life factors can increase the chances? 2) Environmental factors - What are the tolerances of this new technology? Will it be as robust (or more) than flash when it comes to tolerances? 3) Verified data transfers - How accurate are the transfers? It may be possible for it to write & read that quick, but what error correction is going to be necessary to validate the written or read data (like we have for hard drives & flash)? 4) Heat output - Hopefully there's no issue here, but what is the heat production on such a device? I'm glad that the options for future solid state storage are growing, but to echo other's comments it would be nice to see them here sooner rather than later. Like so many other things, it's when we've reached the limits that we'll probably see the next generation "pop up". Cheers

Smiling_eyes_guy report abuse vote down vote up Votes: +0

Re: Self Assembling

Sep 19, 2007 04:31     

Did you read the article even a little bit? Did you pause to think about what you just said? Organisms that react to a natural environment by coating themselves in metal, then moving about to connect themselves to very precise points in a chip on a mass production scale? Come on, don't be stupid. The smallest virus is about 10nm, and all they are is really a neat method of genetic material transfer, they don't even reproduce themselves. Now, on the other hand, if you read the article, well, just read the article... --> Other than current silicon manufacturing techniques that involved lithography, these nanowires, roughly 100 atoms in diameter, use self-assembly, ****a process by which chemical reactants crystallize at lower temperatures mediated by nanoscale metal catalysts to spontaneously form nanowires***, the scientists said.

Bud report abuse vote down vote up Votes: +0

8-10 years?

Sep 20, 2007 08:31     

Recall that 8-10 years ago: a 2 gig hard drive was mainstream, it took two hours to rip and encode a music CD, and usb flash drives were still in the lab. Now, hard drive capacity, memory bandwidth, and encoding times have been reduced by a factor of 50. And we have a new storage medium. So 8 years from now, we should have 50TB hard drives, 500TB/s memory bandwidth, and the ability to encode a 1080p movie in 5 minutes!

Shadowmaster625 report abuse vote down vote up Votes: +0

Not Really New..... if I recall correctly, Phase C...

Sep 20, 2007 11:28     

Not Really New..... if I recall correctly, Phase Change Memory technology was developed by Ovonyx. This "new" variant of the technology by Univ. Philidelphia is sort of a FUD article...the credit and real news was years ago per the technology developed by Ovonyx. See link: http://ovonyx.com/technology/ http://ovonyx.com/corporate/
dave viets report abuse vote down vote up Votes: +0

reader

Sep 20, 2007 11:35     

NOT NEW... . This "new" variant of phase change memory being reported by Univ. Philadelphia is sort of a FUD article. This technology was developed years ago by a company called OVONYX..it is not new. Ovonyx is the same company that developed the CD read/write technology in use today. They have been around a while. See links: http://ovonyx.com/corporate/ http://ovonyx.com/technology/technical-presentation.html

dave viets report abuse vote down vote up Votes: +0

Re: 'self assembling'

Sep 20, 2007 14:56     

Bud is right. Nanowires are mostly grown chemically, unlike quantum dots that are also self-assembled nanostructures grown with molecular beam epitaxy (mbe). The biggest drawback of all those self-assembled structures is the stochastic process of the self-assembly, meaning, that the locations of these structures is random across the growth surface. This is still the greatest advantage of photolithography - the process that site selectively "places" the transistors on current chips and wires them together. There are current attempts in progress for a site selective growth - at least for mbe grown structures, however for chemically grown nanowires it is way harder to obtain such a feature (to my knowledge), that is needed for mass production of future devices. Just suppose you have to locate trillions (=terabit) of these tiny structures on a chip with a microscope and wire them together manually (because the wire distribution is different for EVERY chip!!!)- its kind of unfeasable ;-)And I doubt, that it will be in 8 to 10 years as well! If you guys are interested, just check wikipedia and look up "nanowire" or "quantum dot"

tom report abuse vote down vote up Votes: +0