Memristors combine memory, logics set in one device | Hacker News Reader

Memristors combine memory, logics set in one device

arstechnica.com

24 pointsglymor13 comments

More exciting than this: they've figure out how to use a separate layer of memristors grown on top of conventional silicon logic as a reconfigurable routing layer for FPGAs. This can improve their density, power consumption, and speed.

http://nextbigfuture.com/2010/04/memristor-cmos-hybrid-using...

Another decent use, embodying synaptic weights in neuromorphic engineering. A decent description: http://www.scidacreview.org/0804/html/hardware.html

I think eventually, if we want software that performs like a human mind, we'll want hardware that performs like a brain. That's probably the easiest way to get to a machine that can match us while consuming 30 Watts.

Pretty cool but it seems this isn't new research (there are many references online older than April for memristors and NAND gates).

Can anyone who has access to the actual article explain in a little more layman terms how the ciruit works? I get we have 3 memristors (one of which is always false) and 1 resistor but I don't know enough electronics to understand how they come together and produce a NAND gate that also maintains state.

I do not have access to the original article. But as I interpret this a single memristor acts as a memory device. A memristor has inherent memory. Also combination of memristors may perform logical operations. Thereby you may in theory build computational circuitry from memristors alone. One basic device do it all. That's the point, as I read it.

Usual basic transistor devices are NAND functions

  L L | H
  L H | H
  H L | H
  H H | L

and you need two NANDs for a basic memory cell.

You need only one memristor for a basic memory cell.

Memristors may be combined into IMP functions

  L L | H
  L H | H
  H L | L
  H H | H

IMPs together with NOTs (the 'false' memristor mentioned?) can serve as a basis for everything just as NANDs can.

As I said, this is just my understanding from the post.

This is a tech that was built a long time ago and still hasn't found a use. Also, my boss thinks they're the next computer revolution. This is a man who believes that the asteroid Apophis will probably hit the earth in 2036 and that the Sun is part of a binary system with another star within the Oort cloud. So, I'm pretty sure Memristors are not going anywhere.

While we now believe that Apophis will miss in 2036, it is the large asteroid that has had the largest computed probability of hitting the Earth. So your boss is not far wrong there, and quite possibly right if you're misquoting him in line with your misreading of the article at hand.

And about the binary star system, have you actually read up on the Nemesis theory? While Nemesis has never been found, there is quite a bit of circumstantial evidence for something causing periodic asteroid impacts. So on this issue it looks like your boss is better informed than you are.

Therefore from what you've said I'd be more inclined to take another look at memristors.

2008 was the first memristor; it was only theoretical before that. It says this in the first sentence of the article.

I can't tell if bringing up your boss is satire, it does strike me that way. But if it's not, it's not relevant.

OK, I guess I was wrong about the dates. I must have glossed over that sentence.

The point about my boss is that the things he gets excited about, while they do sort of make sense, have been proven untrue by further science. A correlation test, sure. But still a useful test.

Further, memristors can do logic and memory, but we have things that can do those. They show no advantages over CPU/RAM. I think the only interesting part is the analogy to neurons, but neurons are able to form connections to other neurons and have other machinery as well. Memristors are wholly unimpressive to me.

The big thing about memristors is that they are stable. Think ram that doesn't go blank when you turn the computer off. The first application that comes to mind is basically save-states. You could just shut your computer off, and when you turn it back on, it would resume from where it left off.

Of course, the memristors would have to become much, much faster.

The way I see it, memristors could eventually displace hard-drives, and unify computer memory.

The fact that you mentioned that memristors were "built a long time ago", when they were first built in 2008, clearly shows you are an outsider who has no clue what he's talking about. I suggest you pick some elementary books on electronics and solid state physics and that you abstain from posting your ignorant rants on HN until you have something remotely insightful to add.

The following articles would be a good starting point:

http://www.spectrum.ieee.org/print/7024

http://www.technologyreview.com/Nanotech/20718

http://www.technologyreview.com/computing/21710

http://arstechnica.com/hardware/news/2009/01/logic-circuits-...

Now that daddy is back from the kitchen, it's time for you to leave daddy's computer and go back to your crib to play with your silly putty.

I value the overall content of your comment, but your tone is downright ugly.

You can't predict where technology will be in 10 years. Anyone who claims that fundamental research in some field is useless and / or that it will never go anywhere is... well... a fool. History has taught us again and again that the purpose of futurism is to make astrology look respectable (paraphrasing Galbraith). The fact that HN upvotes trolls such as the one in question is worrying. There should be zero tolerance for self-important brag-a-lots. Extraordinary claims require extraordinary evidence.

Last but not least, using one's boss to make a point, without providing any context or background is utterly sophomoric. Kids these days...

I hope you have a nice life.

Reply on news.ycombinator.com