Slashdot Mirror
Stanford, IBM Team To Explore Spintronics
saxylife writes "NYTimes and various other media are carrying a story on the latest venture between IBM and Stanford," which will concentrate on spintronics, in other words, controlling "the magnetic orientation of atoms to store data.
It's supposed to ease the pressure of hitting the barrier of Moore's law."
Electrons can tunnel across a gate: can variables like spin do the same thing? If so, that's another barrier.
I ope soon we will see a breakthru on media that will require no moving parts in the media, but still give the same I/O speed as current mechanical devices. I know from experience that at least half the time of a drive failure is due to mechanics. But much of the other half is still due to mechanics but appears to be a platter problem?
What this sounds like is a form of bubble memory, a "miracle" technology that was going to take over the world back in the day.
There were actully commercial parts made. But somebody killed it with their idea to have battery cmos ram. Then eeprom and flash memory came along.
They could actually make this work better with the refined manufacturing processes we have today. So I would not discount it out of hand.
Take the cheese to sickbay, the doctor should see it as soon as possible - B'Elanna Torres, "Learning Curve"
You're doing the same thing with 'traditional' electronics anyway. As things scale smaller and smaller, eventually the charge of a single electron will be the limiting factor within a bit, and even before that level is reached, fluctuations of several electrons could be large enough to cause things to "go awry" as you say.
The whole point of spintronics (or magnetoelectronics, it's less buzzword-trendy name) is to add an extra degree of freedom to electronics. Ie, instead of using components that switch on spin-independent electronic charge, one is now adding this extra component that can be switched/amplified/etc.
It's effectively opening up whole new doors, and spintronics represents the 2nd-rapidest movement of technology from lab to market (after the transistor, of course). The field is in its infancy right now, but has huge potential to revolutionize the types of electronic components that exist.
As you say, working on such nanoscale systems makes things really hard, and we're trying now to study and overcome these technical difficulties. But people are hopeful this will produce interesting devices, such as using the spin up/down eigenstates of the electron as the basis states for qubits in quantum computers, for example. Or many other quantum-dependent phenomena that are effectively averaged-out in standard electronics.
make world, not war
It seems that the upper edge of PC power had hit a plateau a couple years ago. I remember 3GHz chips in Summer 2002, a year and a half later, it's now 3.4GHz / 3400+, not the 6000 range that it might have been had the "law" held true.
The parent is modded as funny but it's true. As long as the errors are evenly distributed (i.e., a fundamental effect, vs. a huge-ass scratch across the platter which isn't evenly distributed at all), you can throw enough error correction at the medium to make the unreliability go away.
This isn't theoretical at all; CDs are routinely "destroyed", but they have a lot of error correction built into them so you don't even notice.
Computing the exact probabilities left as an exercise to the reader... but given any level of reliability there is some error correction scheme that can bring it up to any other given level of reliability (short of perfect, of course). Of course you can construct pathological cases that need as many bits as you like, the equations work that way too.
Tiny nano-scale structures change state when they are hit by alpha particles. Consider experimental atom-wide transistors that switch on a single electron. When an alpha particle hits the gate of such a transistor, it flips state momentaily, causing a chain reaction of corrupted data.
Fault intolerance constrains the minimum size of the transistors. There is indeed a maximum speed at which computation can proceed because you cannot continuously shrink transistors in the hope of increasing the clock rates (while maintaining reasonble power). Spintronics offers no solution.
Life is contrained by 5 dimensions: x, y, z, time, and computational speed.