Slashdot Mirror
HP Advances Next-Gen Memory Technology
angry tapir writes "HP scientists have made a small breakthrough in the development of a next-generation memory technology called memristors, which some see as a potential replacement for today's widely used flash and DRAM technologies. In a paper to be published today in the journal Nanotechnology, scientists report that they have mapped out the basic chemistry and structure of what happens inside a memristor during its electrical operation."
[smacks forehead and groans]
If by "about" you mean "about ten times smaller than".
Here's a link (paywall) to the research paper and a free preprint, if anyone cares to read. These *** news sites are never able to publish a link to the original paper.
The only real difference between FRAM and memristors, is FRAM has been licensing and shipping COTS product for over a decade, and memristors are a vaporware product from extremely deep pockets trying to bite a piece off that (admittedly very tiny) market by skating as close as possible to existing patents / copyrights / trademarks without actually being sued out of existence. Its kind of like asking what is the technical difference between a "turbo-" marketed product vs a "i-" marketed product.
Both are basically microscopic core memories. Magnetic field hysteresis, measure magnetic state by trying to force to a given state and seeing how much power it takes, none means its already that state and a bunch means it was the other state.
There are other theoretical uses for memristors. The killer is both devices are current mode devices, which means they'll almost certainly never be power-competitive with voltage mode devices. The other killer is they are not silicon, so that means scrap all the existing fabs and start over. Plus virtually everything out there is silicon based, so it'll be interesting seeing the hybrid devices. And the fourth killer is that memristor/fram technology is advancing, but mass produced silicon dram is also advancing, in fact for a decade or so has been advancing faster, making "modern core memory" ever less interesting.
On the other hand, depending on their temperature handling properties, a memristor based CPU that glows dull red with heat might be OK, don't know. I do know that off the shelf silicon for a variety of reasons doesn't "like" working above a couple hundred degrees, but memristors might not have the same limitations.
"Science flies us to the moon. Religion flies us into buildings." - Victor Stenger
The killer is both devices are current mode devices, which means they'll almost certainly never be power-competitive with voltage mode devices
you have no way of knowing that at this point in time. it doesn't matter that it's a current operated device , it's power that counts.
if it takes 1nA @ 1V to enable/disable a memrister then a billion of them will cost you a watt. So what exactly would be the problem ? Based on my understanding of the physics I think it's completely possible that a memrister could be power consumption competitive with standard technologies.
However your point about the fact that it can't be fabricated using silicon-based technology is a good one, but the memrister seems to me to be simpler to fabricate than a FRAM cell, so that may end up being a wash.
Generally speaking I expect much of the PR today to be vaporware promoting, but the memrister is too new at this point. There might yet be something there.
Absolute statements are never true
The killer is both devices are current mode devices, which means they'll almost certainly never be power-competitive with voltage mode devices.
Are there any voltage mode devices anymore? Flash and DRAM are current mode, and my understanding is that even SRAM uses current mode sensing. I don't work in the FRAM group (I'm a flash guy), but TI's FRAM MCUs are supposed to be super low-power.
CMOS processes always have to be modified to support nonvolatile memory. Not sure what extra steps memristors would need but I doubt it would involve throwing out all existing equipment -- more like adding one or two things.
Visit the