Slashdot Mirror
Advance In PCM Memory Could Dramatically Reduce Power Consumption
Zothecula writes "Researchers from the Electrical and Computer Engineering Department of the University of Illinois have developed new low-power digital memory which uses much less power and is faster than other solutions currently available. The breakthrough could give future consumer devices like smartphones and laptops a much longer battery life, but might also benefit equipment used in telecommunications, science or by the military."
Five moderator points, and no comments worth reading.
Orwell: "In a Time of Universal Deceit, telling the Truth is a Revolutionary Act"
Two questions that hopefully more technically knowledgeable people can comment on. First, this system uses nanotubes- as I understand it, there's no good way of making high-quality naontubes in large batches. Is that still accurate? Second, for most devices isn't the CPU consuming a lot more energy than the memory? If that is the case, won't more efficient memory have only a small impact on overall power consumption?
Noticeably muddier and harsher than DSD memory, though.
How can I believe you when you tell me what I don't want to hear?
Abstract
Phase-change materials (PCMs) are promising candidates for nonvolatile data storage and reconfigurable electronics, but high programming currents have presented a challenge to realize low power operation. We controlled PCM bits with single-wall and small-diameter multi-wall carbon nanotubes. This configuration achieves programming currents as low as 0.5 A (SET) and 5 A (RESET), two orders of magnitude lower than state-of-the-art devices. Pulsed measurements enable memory switching with very low energy consumption. Analysis of over 100 devices finds that the programming voltage and energy are highly scalable, and could be below 1 V and single femtojoules per bit, respectively.
I didn't read the TFA, but they did use a TLA acronym, so you can GTFO out of here.
Were any Rambus people within earshot? Any patent applications suddenly filed by them?
A feeling of having made the same mistake before: Deja Foobar
TFA explains how to set a bit (from 0 to 1), but not how to clear the bit.
The actual paper in inaccessible to me (without $$$), even from a university, as it is pre-publication. (I think it will become available on the 17th.)
Quattuor res in hoc mundo sanctae sunt: libri, liberi, libertas et liberalitas.
This is the worst /. article *summary* that I've read this week. TFA addresses phase-change materials (PCM)-based memory only. This is the self-same stuff used in writable DVDs, and has some very cool properties.
In the above summary, "Faster than other solutions currently available" refers specifically to PCM-based memory. The durability of PCM memory is one big plus -- all those sci-fi plot twists from cosmic ray induced bit-flips in charge-dependent memory? Yeah, not a problem here.
TFA itself is a really neat little paper. It's in Science, which indicates that some reviewers somewhere thought it both important and well-done. It's surprisingly readable, too, which is a little unusual for these sorts of papers. These folks were clearly thinking about Fab-type high volume / high yield questions. For example, the "quality" of the carbon nanotubes (CNTs) isn't important. They're just an easily-broken conductor. Clearly this isn't ready for prime-time, but they didn't just make 1 device, test it, and publish. They made at least 100, while varying conditions.
From TFA's Supporting Online Material:
" In order to create the CNT nanogaps, we performed electrical breakdown of CNTs both in ambient air and under Ar flow. We have also cut CNTs with AFM manipulation, but the elec- trical breakdowns offered a much faster route to obtain a wide range of nanogaps (Fig. 4). Of course, while the CNT breakdown method is extremely useful here, it would not be the preferred route for obtaining nanogaps in a more scalable manufacturing environment. Nevertheless, we believe it is useful to present some observations associated with this technique here.
First, we note that CNT breakdowns under Ar flow were done by flowing Ar (which is heavier than air) from a small nozzle over the entire test chip while probing. Thus, some dimi- nished amount of oxygen was still available for CNT breakdown, unlike the breakdowns per- formed in vacuum in the second panel of Fig. 2C of Ref. (1). There, the CNT break in vacuum could lead to SiO2 damage, which was not seen here either in ambient air or under Ar flow.
Second, we found that nanogaps formed in Ar are always smaller (always
We report additional statistics for all devices measured by AFM in Fig. S7. We find no clear dependence between nanogap size and CNT diameter (Fig. S7A). In a sense, this is encouraging because it suggests that tight control of CNT electrode diameter may not be necessary to make very low power devices. Our simulations (Figs. S3 and S4) also suggest this is the case, because the resistance of the GST bit always dominates that of the CNT (both in the a- and c-GST phase), thus rendering variability in the CNT of less importance. This fact could be important for mass production of such electronics where some amount of CNT variability could be tolerated."
From the article:
the solution needs to be upscaled. The group has so far created and tested a few hundred bits
It's another one of those overhyped materials-science articles. Here's the hype part:
Pop says that he envisions a point where a device could get its power needs from harvested thermal or mechanical energy or sourced purely from solar. Consumer devices won't be the only beneficiaries, however. "We're not just talking about lightening our pockets or purses," Pop said. "This is also important for anything that has to operate on a battery, such as satellites, telecommunications equipment in remote locations, or any number of scientific and military applications." Server farms or data centers could also benefit from lower energy costs by utilizing the solution. The researchers say that the low-power memory could even lead to previously elusive three-dimensional stacking of chips.
Really. They've come up with yet another alternative to silicon-based memory devices. There are hundreds of such schemes, from Ovonics to silicon-on-sapphire. Many of them work, but each has something that makes it inferior to the mainstream technologies. Some (like this one, probably, since it's heating-based) have slow write times. Some are expensive to fab. Some won't scale up.
The problem with the "elusive 3D stacking of chips" is not that it can't be done, but that it doesn't make systems cheaper. In the technologies developed to date, each new layer of devices costs about as much as making a separate part.
So, everyone said, if we switch to SSDs, longer battery life. Did it happen? No.
Memory does not use up that much power, relative to the whole system. Even switching a laptop's screen to LEDs doesn't help that much.
It's almost the same as the mythical new invention that will be out "in 5 years".
Give it up people. Semiconductors improve year after year, These kinds of breakthroughs that drastically change everything just don't happen.
The only thing I can think of that made that kind of a change in the last 30 years was going to an SSD drive for speed and responsiveness. Other than that, each year gets a bit better. Don't expect that to change any time soon.
Don't steal. The government hates competition.
I, no doubt like most of you, enjoy reading about advancements in technologies. Especially those which can have a direct impact on computer performance!
Lately though, it seems to me, that all of these headlines and articles are exactly the same article. Making some claim about a major breakthrough and how this will make my computer 100 times faster and use 100 times less power....maybe...in the future.
It seems always to the case that in 10 years, "this" will enable us to use a lot less power and have much higher densities and so on and so on.
Basically, why the techs are interesting, the articles are absolute crap. Sensationalizing everything where there is nothing worth do so about. Why can they not just explain the so-called break though, it's probable applications and realistically how long, if at all, until commercialization?
Actually, the headline reads 'Advances in Phase-Change Material Memory...' if you want to expand the initialism. PCRAM is the more common abbreviation for phase change (random access) memory. It's a very interesting technology for persistent storage, for two reasons. The first is that, like DRAM, it's byte-addressable. The second is that read speeds are almost as fast as DRAM. This means that you can just map a bit of it into a process's address space, without any copying, to make the data available. It's a good contender for swap / hibernate space, because you can write pages that haven't been modified recently out to PCRAM, update the page tables to point there, and then turn off DRAM chips. Writing to PCRAM is quite a bit slower than writing to DRAM, but is probably fast enough when you're in power-saving mode, so you can turn off all of the DRAM, and only turn it on when the system load increases again.
I am TheRaven on Soylent News
Depends on the mode. One thing people really hate about mobile devices at the moment is their battery life. The problem with DRAM is that it draws almost the same amount of power, irrespective of the system mode. Every bit in memory is constantly being refreshed, and this draws power. When you put your machine into suspend mode, about the only thing still drawing power is the RAM, and this gets worse the more RAM you put in the machine (which is a big part of the reason why mobile phones have a lot less RAM than laptops, even though 4-8GB of RAM chips would easily fit in a phone). PCRAM, like flash, uses no power when idle, a small amount of power when being read, and a lot more when being written, and is almost as fast to read as DRAM. This means you can do the equivalent of suspend-to-disk / hibernate, but resume almost instantly.
I am TheRaven on Soylent News
make that a memristor sandwich and you would have my interest. in fact i've decided not to buy another pc until it features AMOLED and memristor memory. fuck these other "advances", they're just trying to stall the inevitable or grab some of the remaining market time. toshiba just got quaked and everybody's bitching about how memory's going to go up. someone is inevitably going to attempt to capitalize on that, but unless it's the *actual*, new, advance it's not really "advanced". so opt out of planned obsolescence and hold out for memristor from HP.
"Stratigraphically the origin of agriculture and thermonuclear destruction will appear essentially simultaneous" -- Lee
While these folks may have made something 100x faster than something else, it's still useless.
It's a phase-change -- that means HEAT, which means it takes time for the storage element to heat up, change phase, and then cool down. It's probably better than using punched cards, maybe even better than 1990 flash memory, but not a whole lot better.
Memory technology today has to work on the nanosecond level-- phase-change is not going to get anywhere near that ballpark.
You also need storage that can be read and written thousands of times without degradation-- phase-change is unlikely to ever get to that level.
Why wouldn't this mythical device be applicable to medical devices? It seems there are a dozen other industries you left off your list, is it not potentially applicable to them either?
This is just another one of those advances that'll eventually make its way into shipping products, and no one will notice because it will be part of the general trend of improving technology.
For instance, look at all the (relatively) recent breakthroughs that make modern tablets possible:
- Efficient white LED backlights ...there's more, but my point is, there were probably press releases about these years and years before they made it into shipping products, and we completely forgot about them. We just saw an iPad and went "Ooh that's cool!" without realizing the advances made years before which made it possible.
- Li-Polymer batteries
- High density flash memory
- Software defined radios
- Capacitive touchscreens
I was looking forward to new memory that might be more expensive, but has the best of all worlds.
(simple hardware interface, durability and faster than dram write and read speed, ie. like sram but cheaper and only uses energy while active)
The embedded world seriously needs new parts that allows designs and software to become simpler, at the expense of hardware cost.
Well, this is Slashdot. People here saw the iPad and poopooed it because a) it's Apple and/or b) they're super-geniuses and could have created one in their sleep, if they would just lower themselves to something so mundane.
Agreed. And PCM Memory will remove the standby power required for DRAM, which is a significant drain on battery life.
Of course, the jury is still out on whether the writes will take more power than DRAM. If so, the power savings could be much less.
Man is the animal that laughs.
And occasionally whores for Karma.