HP's "Memristor" Could Replace Transistors

Page 2 - Seeking answers? Join the Tom's Hardware community: where nearly two million members share solutions and discuss the latest tech.
Status
Not open for further replies.
[citation][nom]loomis86[/nom]I'm having a real difficult time imagining software that can make use of a CPU (or RAM) composed of switches with more than two possible values.[/citation]
As you should. As we all should. This means more than just smaller, faster, computers. This means that computers no longer have to be digital. They can essentially be analog, like us.
 
Helios.

No, if this tech DOESN'T get covered up for 50 years as to not destabilize the current status quo of the tech market it could lead to something cool. Call me a skeptic, but so much good tech has been buried for financial reasons in the last decades of technological improvement that it's not worth mentioning.
 
[citation][nom]dreamphantom_1977[/nom]This is old news- "Published: May 1, 2008". I knew I read this somewhere before, scientific america I think. When I clicked on the link they posted, it shows the publishing date right at the top of each page. Proving my theory that Toms is running out of articles. Starting to look through old news. LOL.[/citation]
Why was this guy voted down? He's right. The least that could have been done is a little research and including info from the BBC article.
http://news.bbc.co.uk/2/hi/technology/8609885.stm
 
[citation][nom]FUtomNOreg[/nom]Speech recognition and AI has never been a problem of technology. No amount of speed or capacity can compensate for the simple fact that science still does not understand how our minds "work." Both require the understanding of context; it's not just an exercise in brute force resolution of pattern matches and decision trees. This claim gets rehashed every couple of years and will continue to do so for the foreseeable future.[/citation]

the ability to have speech sound more realistic is what this is referring to... instead of the semi-monotone computer voices you hear now.... they have gotten better over the past few years but you can always tell the difference between a human and a computer speech... this opens the possibilities for realistic speech...

as far as vision.... instead of seeing "billions" or "trillions" of colors... like our cameras do now... imagine computers actually being able to see everything clearly as you and i do...

this is only saying its a step towards true AI... specifically in those two areas... i agree with you on the understanding the human mind is a task we are nowhere near completing... but this opens more doors...
 
But can it win crysis?

Seriously, though. They talk about these things as possible components of neural networks. But until you have the ability to connect each element to lots of other elements, and alter the "strength" and polarity of each of those connections, you don't have the ability to make such a network. The memristor could provide for the variable connection strength, but more research is needed to achieve the level of interconnectedness that is required.
 
this is pretty amazing as it could bring back semi analog computers.

for graphics cards they could transmit 32bit RGBA info as 4-bit analog info resulting in simpler architecture which allows for more graphics pipelines and less power consumption.
 
[citation][nom]husker[/nom]As you should. As we all should. This means more than just smaller, faster, computers. This means that computers no longer have to be digital. They can essentially be analog, like us.[/citation]

ok, so what software can do that? It doesn't exist, that I know of.
 
[citation][nom]BPT747[/nom]Nothing like a vague description, I was looking to see a structural diagram with doping levels and their material construction that can sustain saturation when no base current is flowing. Unless this is just another name for Zram(sounds like it anyways) technology that I read about 3-4 years ago in which this is not really a new concept and just a combination of a transistor and a capacitor as a single device and doesn't retain memory when off.[/citation]
You have no idea what you are talking about on this subject. True, this article isn't very clear about the subject matter, but you should go to HP's website and look it up. They've been working on the technology for a couple of years, but memristors were first postulated on since the early 1970's--originally at Berkeley.
 
[citation][nom]enzo matrix[/nom]Appearently "rather quick" is still much longer than 10 years. Either that or the technology was abandoned. Ideas anyone?[/citation]
It is being worked on. There is a project at DARPA with IBM, HP and a few other companies involved to create adaptable neural networking systems. There's been rare tidbits of this project in the tech news circles here and there, but it exists.

Electronic circuitry has always required a pre-planned current state--meaning it could only accept a certain voltage--basically passive electronics. The memristor will allow active electronics where you could change the voltage amount on the fly in electronic systems--this is revolutionary as a device could do more than one thing. Also, instead of just having a 1 or 0 voltage state for logic purposes, where you have if/else scenarios, you can have 0, .1, .2 volts etc... So, instead of building a binary system, you could say, build a human-relational decimal system with 10 states--just as an example of one possibility. This greatly expands the logic possibilities and would actually make the system easier to program, instead of harder as some people have incorrectly stated. Is it harder to program a robot to do something or harder to program a dolphin to do something? Obviously, the dolphin is easier to train and get to do things as the military has proven with things like underwater mine spotting. It's the same principle. Though it may be counter-intuitive to some, the more robust system would be easier to deal with.

Aside from CPU's and circuitry, the memristor is also ideal for memory applications. This wasn't really explained in the article either. Here is a good source on the subject.

http://www.eetimes.com/news/latest/showArticle.jhtml?articleID=224202453

Memristors will be used militarily first in various electronics and will then work their way toward the consumer--probably toward the end of the decade.
 
Our dependence on the transistor, an ancient (by today's standards) invention, is what is slowing down technological advances. In particular the MOSFET. It's a power hog so you can't do much with it. If power consumption was a few orders of magnitude less then it would be much more useful.
 
[citation][nom]stevo777[/nom]It is being worked on. There is a project at DARPA with IBM, HP and a few other companies involved to create adaptable neural networking systems. There's been rare tidbits of this project in the tech news circles here and there, but it exists. Electronic circuitry has always required a pre-planned current state--meaning it could only accept a certain voltage--basically passive electronics. The memristor will allow active electronics where you could change the voltage amount on the fly in electronic systems--this is revolutionary as a device could do more than one thing. Also, instead of just having a 1 or 0 voltage state for logic purposes, where you have if/else scenarios, you can have 0, .1, .2 volts etc... So, instead of building a binary system, you could say, build a human-relational decimal system with 10 states--just as an example of one possibility. This greatly expands the logic possibilities and would actually make the system easier to program, instead of harder as some people have incorrectly stated. Is it harder to program a robot to do something or harder to program a dolphin to do something? Obviously, the dolphin is easier to train and get to do things as the military has proven with things like underwater mine spotting. It's the same principle. Though it may be counter-intuitive to some, the more robust system would be easier to deal with.Aside from CPU's and circuitry, the memristor is also ideal for memory applications. This wasn't really explained in the article either. Here is a good source on the subject.http://www.eetimes.com/news/latest [...] =224202453Memristors will be used militarily first in various electronics and will then work their way toward the consumer--probably toward the end of the decade.[/citation]

no one writes software for a dolphin. The dolphin comes with with the operating system pre loaded...as well as all drivers for all peripherals.
 
[citation][nom]loomis86[/nom]ok, so what software can do that? It doesn't exist, that I know of.[/citation]
It does, but we don't know how to copy or download it yet. In fact I'm using it to type my comment, just as you are using it now to read it.
 
[citation][nom]loomis86[/nom]no one writes software for a dolphin. The dolphin comes with with the operating system pre loaded...as well as all drivers for all peripherals.[/citation]

Exactly my point. It learns on its own. That is what the promise of the memristor holds.
 
Status
Not open for further replies.