Slashdot Mirror
HP Patents Nanoscale "Street Map" Technology
radsoft, pointing to this HP press release, writes: "HP says silicon electronics will reach a dead end in 2012, and wants to have a 16KB prototype of its molecular memory working by 2005." Basically, it looks like they've worked out some of the details of interfacing molecular components (still in their infancy of course, but promising) with traditional silicon.
How does slashdot differentiate between good patents and evil, bad patents? Is the litmus test "Jesus Christ, I could have thought of that!"?
- A.P.
--
"Remember when the U.S. had a drug problem, and then we declared a War On Drugs, and now you can't buy drugs anymore?"
I'm really enthused to see that DARPA is funding ~half of this project, but HP gets to hold the patents. Maybe we need a million geek march on Washington to tell them that if we are paying for public (not "national security" related) research (through tax dollars), we expect to have rights to said developments.
Something ironic in there about the government funding research so that we can be forced to pay a company for it.
Here's a link to an article in NYT that has more details than the press release.
--
"Man in the Moon and other weird things" - wfmh.org.pl/thorgal/Moon/
"The world's petroleum reserves will be depleated by 1980."
"The world demand for computers should be no more than 5 or 6 units."
"No one should ever need more than 640KB."
"Silicon will be a dead technology by 2012"
Umm, is it just me, or am I the only one that sees a pattern here? For christ's sake, the only thing that will probably die (and SHOULD die) are uniprocessor systems. When we get to the point where SMP is an obsolete technology, let me know. Until then, theres absolutely no need to push in a new direction when it comes to the way we fab processors..Time and time again, history has shown that the instant ANYONE whips out a hammer and chisel in this industry, and starts making a tombstone for any technology, they're views ultimately go down the tubes as alarmist and horribly myopic.
My $0.02,
Bowie J. Poag
HP is taking the "kill em all and let God sort em out" approach here, by using some sort of chemical etching process that makes a rat's nest of random possible connections, then figuring out which connection goes where after it's over. A couple of questions arise...Isn't there a possibility that one or a few connections just don't get made and so the circuit just won't work, even tho it's 99%+ connected? How often will something like this happen, and will it make the wafer yield too low to be feasible? How will you certify something like this for NASA and the military - they already are a little leery of things like neural nets which aren't deterministic enough to fully trust in mission critical aps. And finally, after a while don;t you get so small that cosmic rays / radiation will zap the wires? Transient resets in CPUs from cosmic rays is already a measurable phenomenon, would'nt this be worse?
This is true; however we certainly have yet to exploit all the possibilities when it comes to heat control. Processor voltages continue to drop. In your example a 100-layer CPU would have 1000W of disapation. However, it would also have 5 times the effective tranmissive area, so in theory we could build a heatsink that is just 5 times the size of modern CPU sinks and dissapate fully half that energy. (provided the ambient air is kept cool enough) If the voltage drops by 1/root(2) we have a feasible design.
Big-O means we'll always be limited in the 3d growth of chips, unless practical superconducting ICs come along. However, the problem is really not surface area as much as total energy output. I doubt that the average consumer wants a 1000W cpu of any kind in their box; at that disapation manufacturer's need to worry about burning down their customer's houses, to say nothing of the electrical cost. In addition, the forced-air heatsink would need 100 times more surface area or much more air; pretty prohibitive!
Within reasonable limits, we can generally get rid of heat in proportion to the linear size of the CPU, especially when you go to water (or better) cooling, increase the surface area by boring passages through the chip itself, etc. Even when the chip becomes so small that metallic sinks are ineffective at transmission we can always immerse the sinks in the cooling medium. Only then will we approach your hard surface area vs. volume limit.
Only a few decades ago, having that much memory would take the size of small room, and the money of a small country, in just a couple of years (2005) these guys are going to have it at molecular size (admittadly, still with the money of a small country no doubt) !
I just can't wait for usefull nanobots to become a reality.
NZ Electronics Enthusiasts: Check out my Trade Me Listings
It's worth thinking about the history of aircraft development. Aircraft performance showed a steady, rapid growth curve from 1900 to the late 1960s. The 1960s produced the SR-71, the 747, the C-5A, and the Concorde. And there things stopped. All those aircraft are still flying, and performance improvements since that era have been marginal.
In the 1960s, the aviation industry expected continued rapid progress. Supersonic and hypersonic transports, spaceplanes, suborbital ballistic transports, and rocket planes were expected in the near future. Super fuels with higher energy density were discussed. Giant aircraft capable of carrying many truck-sized shipping containers were proposed. Nuclear powered aircraft were in development. There was serious DoD-funded work on antigravity.
None of it worked. Aviation never moved to a new technology. Wings and petroleum-based fuels were used by the Wright Brothers, and that's what we use today.
This could be what happens to electronics. There are several known alternatives to optical lithography on silicon. But none of them are better in a business sense.
True. And you can build up multi-layer ICs. Slowly. The problem is that fab time, and cost, climbs linearly with the number of layers. So there's no real benefit to doing it other than packaging density. And, of course, there's the yield problem.
There are some semi-3D technologies. One consisted of making up a stack of dies with connections along one edge, slotted into a "backplane" wafer like a rack of boards in a card cage. The problem is that at current RAM speeds, getting rid of the heat is a bigger problem than finding physical space for the memory modules. For a while, there was enthusiasm for wafer-scale integration. Think of a whole 6" wafer of RAM as a single component. Products were demonstrated. But there was no point; traditional dicing, packaging, and assembly into memory modules turned out to be cheaper than packaging a giant IC, something that required a heavy metal frame to keep the chip rigid.
I always liked direct-writing E-beam fab technology myself. It's been tried repeatedly for about 25 years now. Works fine, writes slowly, costs more.
The problem is not that there aren't higher-density technologies than lithography on silicon. A number of approaches are known. But they're all more expensive.
There's an analogy with supersonic flight. Works fine, and inherently costs about 3x more than subsonic flight. That's enough to kill the commercial market for it.
There may be future 3D technologies, but they'll probably be fabricated by something that looks like biological growth or real nanotechnology, and can thus work on the whole volume simultaneously. That's a long way off.
What a smashing way to turn one's boobs into a computational device.
If you open yourself to the foo, You and foo become one.
Is it just me, or does it seem you'd still be limited to the speed of existing micro technology?
---
Video meliora proboque deteriora sequor - Ovidius
Two kilobytes? WOW!
Why, if Moore's law applies to this new technology and they get a 64-fold increase over the following decade, they'll have built a 128KB memory by 2015!
Move over, DRAM! Step aside, SRAM! A new memory king is coming to town!
Tim
Actually, 16 kilobits. That's some sweet power.
Fast forward, 2005. Buy one of these nano machines from Target. Buy copies of COMPUTE! NANO. Stay up till 3AM entering programs into it like this:
25,254,3,5,32,60,251,232:4A5D
...
Remind me again how patents are supposed to foster innovation and benefit the economy?
Sure thing. If you invented a device that cures cancer, patches the hole in the ozone layer, and cooks a juicy turkey in under 5 minutes, there's not a company in the world who would touch it if you didn't have a patent. The reason is that without a patent, the company would almost certainly *LOSE* money on the device. Getting a successful product to market requires not just the initial invention, but product development, packaging, market research, advertising, product placement, movie tie-ins, etc. Each of these things requires time, money, and people (i.e., money, money, and more money).
When company A finally releases the product, the list price is significantly higher than the manufacturing cost, because company A has to recoup the costs I just listed. However, if the technology isn't protected by a patent, then company B can come along and release an identical product for only slightly more than the manufacturing cost (since company B didn't have to do any of the ancillary work), undercutting company A and usurping all of the income for the product.
So without a patent, company A does all the work and company B gets all the profit. Clearly, company A isn't going to engage in this kind of business. However, if the invention is protected by a patent, then company B can't undercut company A's prices (or at least, company B has to pay company A for using the technology), so company A can make money by developing the technology. That's how patents foster innovation and benefit the economy.
Go Lance Armstrong!
On stereophonic equipment, the monaural sound obtained through multiple channels will enhance your listening pleasure.
"16K ought to be enough for anybody."
:)
"Your superior intellect is no match for our puny weapons!"
Having read the article i cannot see any actual real breakthrough at this stage - what they are saying is that molecular memory may be possible - this is not a new theory.
Their 'patent' by the looks of it is being a protective measure in case someone else comes up with the idea and actually gets it to work - they estimate they will have a 16kb prototype in approx 4 years - in other words they have patented the theory that this may work but now one has actually physically made one yet - they are working on this but just in case someone else manages it HP will have the patent and thus are due a royalty on anmy product ?
Also one more comment - this is really nothing that unusual - patents like this are awarded all the time - yet this is getting some news covergae - the wording makes it sound at first glance like they have this thing working, only when you read it do you realise they dont.
Arent H P in some trouble at the moment - their profit is down and several of their divisions arent permforming and their stock price has fallen, their is talk of managment shuffles at the top. Could this be a bit of positive news to salve the market ahead of the realease of their financial data today ?
I refuse to argue with Anonymous Cowards - if you want a discussion get an account....