Slashdot Mirror
UCLA Chemists Progress Toward Molecular Computers
Concepvelo writes "It is very refeshing to see Professor Stoddart (my organic chem professor last quarter) and Pat Collier making progress toward molecular computers. Stoddart's team has created molecules that can be switched hundreds of times, where before they could only be switched once. They are saying that the creation of Molecular RAM is one step closer because all of this can be done at room temp. The article is here."
And this doesn't happen to existing computers? We're putting so much energy into such a little space that the heat is phenomenal. 50 watts of heat over an tiny little 50mm x 50mm cube? That's incredible. If the case temperature rises above about 90 or so, most systems get /very/ unstable. Most modern computers are (or should be!) kept below room temperature - at around 68-72.
At the rate we are accelerating at, radical new solutions will need to be devised to keep up with heat output of newer chips.
Personally, I think a molecular computer would be easier to manage than our own. If it were possible to assemble a system that could function at -20C, I would try to get on-board their project to help with the HVAC equipment... it would be more energy efficient in the long run.
" ... a realistic memory cell needs to switch billions, if not trillions of times during its expected life."
yes but 1 compared to 100 is 10^2. We're talking about advancing technology exponentially.
I don't think it will be long before these are useful. A digital camera for example would be able to make use of memory chips with that existing technology. 300 Picture Packs of Ram -- once a process is introduced and prices come down, it may not even be necessary to improve the technology - though I'm sure it will be.
Ace
These are awesome developments for both molecular engineering and the high tech. Remember, though, that many new developments never see commercialization, because they just can't be made faster, cheaper, better. There are so many dead technologies that were promising once. Memory such as 3d protein gels (read and written with lasers), holograms and others just haven't seen their full potential, and possibly never will.
-- Moondog
The site is bad. Check out a mirror over at this site
/. is a commercial entity. goto slashdot.com
I think they are talking about HP's 'learning' technology which allows them to take chips with 'dead' portions and have the logic discover those dead parts and work around them. I think it was called something like the Terramac? Saw an article a few years ago about HP putting a bunch of their broken/out-of-spec PA processors into this machine, and then a piece of software mapped out regions that had problems. This would fundamentally help molecular computing as molecular-level manipulations are going to be sensitive to things such as cosmic radiation pushing molecules into undesirable states. Software that can route instructions around such problems will be a cornerstone for such technology development. Also I believe the purpose of pairing these two technologies was to achieve some sort of 'shake-n-bake' processor development. ie, put in a bunch of switches and carbon nano-tube 'wires'.. shake it up, and then use this piece of software to discover a way for processing to take place. This would allow them to develop the technology without requiring cheap nano-manipulation. Other attempts at this are being made with self-assembling systems/organic systems. Even if this technology was perfected and cost effective now, I don't think it's in the interest of *any* of the powers that be to release such a disruptive technology immediately. This would completely gut the chip industry.... So who knows.
Actually, if the memory is fast enough, you could use individual memory bits as logic gates quite easily - which means that the only thing for a computer which WOULDN'T be memory, will be the connections inside the chip & the connections to the external world.
You're just lucky, not better at doing anything... I bought a P3-600E recently.. When I got it setup I noticed the temperature was in the 80s (Celcius) from the on-die sensor. This was without tweaking, with the regular fan, running properly, in an open case. This was *way* too hot, the CPU warning sensor from the motherboard went off every now and then which didn't make me feel it was terribly safe. So I went out and got a good fan, an Alpha PEP66, then the temperature rarely went over 38C, which is high, but is fairly normal considering the on-die sensors are a bit higher. So I overclocked... My CPU wasn't stable with the retail Intel fan/heatsink at its rated speed. But with a good cooler, not only was it stable at its rated speed, it was able to run at 800... I actually got it a bit faster but dropped it back to keep everything in spec considering I've got some cheap PCI cards. :(
So, your stuff isn't overclocking just because its all older and doesn't draw as much power into as small an area, not because your reluctance to overclock actually helps it.
Pulling my head out of my butt, let me correct myself: Currently CPUs are on the order of ten times as fast as the MEMORY.
You cannot apply a technological solution to a sociological problem. (Edwards' Law)
Yes, they are an order of magnitude or two slower, but they aren't a critical, first-order resource. Disks are "secondary storage," remember. (I know it sounds weird to think of them like that, but it's true.)
You cannot apply a technological solution to a sociological problem. (Edwards' Law)
As those of who work with computers know, memory is only part of a computer. Yes, an important part, but only a part. As the article points out, getting these molecular switches to work is only part of memory for that matter. Before we can will be seeing NanoRam (NRAM)TM on the store shelves, they have to figure out how to wire lots of these little beasties together. But it would be nice to have Tera-bytes of power in a quarter sized computer.
Another problem I see with this early version of memory is that the individual cells can be switched "hundreds of times". Better than once, yes, but a realistic memory cell needs to switch billions, if not trillions of times during its expected life. We have a way to go here.
Gonzo
Heh... with the numbers of chipsets and sockets et al that motherboards in circulation today have, we 're almost at that point already. ;)
--
rickf@transpect.SPAM-B-GONE.net (remove the SPAM-B-GONE bit)
"People will pay big bucks for the luxury of ignorance."
The author of the article wrote:
This doesn't necessarily mean that they stop working after that point. It's reasonable to assume that a molecular device that can perform a repetative function can continue to repeat that function.
If I were to say that I've rebooting my computer hundereds of times, you wouldn't assume that I can no longer boot it.
Also, it helps to pay closer heed to the quoted source than the authors text. The author quoted:
Molecules don't break down from wear in quite the same manner as larger scale components. Assuming that such switches are properly housed, there lifespan would be affected by things like changing environmental conditions (EM, temp, etc.) but not repetative use.
The part I objected to is the following:
This just doesn't ring true. First, they are far from actually developing a "molecular computer". How could they be working on one that does anything, let alone a learning one.
This sounds like hype... call it "text-candy" for lack of a better term. The principles for bullding a computer are not really changing here (switches, logic gates, etc.) so developing a "learning molecular computer" has two steps:
They are currently on step one....
Ok, you begged the response:
Only if you were running windows/linux/freebsd/dos/os2.... </joke>
---
-
ping -f 255.255.255.255 # if only
One thing I noticed about the article is that the term "nanotechnology" or "nanocomputers" wasn't mentioned once.
The closer they got was "nano-sized computers" and an institute called nanosystems.
I don't think this is coincidence. After a lot of media coverage of stories like Bill Joy's paper (which i read and liked very much, but didn't totally agree with) the term nanotechnology has been made to have a negative ring to it.
then you have all the nonesense far less inteligent people have said, plus some very cool renditions like Deus Ex, and what you have is instant holocaust. soon nanotechnology will sound as bad as nuclear reactors and genetic engineering to a lot of people (they all sound fine to me).
anyhow, congrats to the research team.
There are two kinds of people in the world: Those with good memory.
This kind of research sounds like a waste of time -- given how fast conventional processors are accelerating (Moore's Law and all), do we even need molecular computing?
Well of course we do. As things get smaller and smaller it's going to get to individual atoms and molecules anyways.
Even if the current silicone transistor technology, can be scaled much further down. It will still take up a lot more molecules than this (possibly in the millions), simply because it's a technology that is not design to work with individual molecules.
My computer is currently not powerful enough to do real time voice and video processing, determine from my tone and facial movements what I want the computer to do, or providing intelligent assistance, or doing even rudimentary AI tasks. All home computers do is flashy graphics - I want the power to HELP me in something IMPORTANT. I want the ability to havea thousands of nano-sized devices implanted in my skin to store critical information. I want my own Personal Area Network (PAN) that is completely secure and stores a record of all my experiences, emotions, health readings, and physical stimulus. I want a complete, real-time searchable, intelligent database system that monitors what I do and prompts me with intelligent and relevent information to help me with the task at hand.
Do we have the processing power to do this? Yes. Could I fit in my pocket? No. That is why this is important - every step towards nano-scale technology is a leap forward towards a future of prevelant, pervasive technology that is much more flexible than what is currently available.
THAT is why this interests me.
Great, now biological and technological warfare can converge :) Nerve gas won't just be for humans! *lol*
I have this theory that with nano, EE/CS will become in less demand, and mechanical engineers will be forced to reexamine rod logic if they want the good jobs. However, we're really moving at a snail's pace here, and haven't had any real developments in a while. nano-saxaphones for Pres. Clinton! ;-)
(Yes, I'll take a couple of those 6.0225x10^23 SIMMs, please...)
--
rickf@transpect.SPAM-B-GONE.net (remove the SPAM-B-GONE bit)
"People will pay big bucks for the luxury of ignorance."
I agree with you, I just want to expand on something:
And it's the part that can benefit the most from this kind of thing. Not for size, but for speed. The speed of memory access is often the controlling factor in how fast a program runs. It doesn't matter if you just bought a 6000 terahertz CPU; if your memory is slow, your processor only waits more.
Currently CPUs are on the order of ten times as fast as the processor, and the gap is increasing. We need faster memory more than we need faster anything else. (Well, maybe faster pizza delivery.)
You cannot apply a technological solution to a sociological problem. (Edwards' Law)
There just seem to be a whole lot of risks involved with molecular computing. Wouldn't it be easy for the molecules to be jarred out of position? If a part of the computer breaks, would all the molecules inside be lost? This kind of research sounds like a waste of time -- given how fast conventional processors are accelerating (Moore's Law and all), do we even need molecular computing?