Whether or not its a breakthrough really depends on its speed. If the new memory is slower than flash, then its not going to be on market. However, if its faster than flash, then we will see some very interesting things in the future.

Mine is the... eh... the... last one when I return just before closing time...

"Another potential application of memristor technology could be the development of computer systems that remember and associate series of events in a manner similar to the way a human brain recognizes patterns. This could substantially improve today’s facial recognition technology, enable security and privacy features that recognize a complex set of biometric features of an authorized person to access personal information, or enable an appliance to learn from experience."*

This is potentially a far more significant application than simple non-volatile memory, for which a number of solutions exist or are being developed.

*See http://www.physorg.com/news128786808.html

Where did the term memristance come from? Resistance is a measure of how a substance resists the flow of electrons, capacitance is a measure of charge holding capacity and inductance is a measure of current induced by mag flux (or vice versa). So what is the root for memristance? Turns out that memristor is derived from memory resistor. A nomenclature that suggests that rather than being a fundamental element it is derived or dependent on other fundamental elements.

What is new about the functionality that Williams claims, seems that HP's creation is just an alternative to something that already exists - flash. Flash is constructed from the 3 basic components and transistors, which can also be defined in terms of the 3 basic components (eg hybrid-pi model). So either the description of the memristor function is inaccurate or it appears that there is no 4th basic circuit component afterall.

A little research seems to suggest that Chua looked at R = V/I, 1/C = V/Q and L = Phi/I and decided that there must be somthing that equals Phi/Q as that combination is missing from the set. This seems to suggest that a memristor will be susceptable to influence by external magnetic fields, something that Williams doesn't appear to mention even though it would have implications for practical applications of a memristor. Or maybe Williams et al haven't actually produced a 'real' memristor...

Sounds cool but just not a new fundamental passive circuit element.

"The memristor consists of two titanium dioxide layers connected to wires. When a current is applied to one, the resistance of the other changes. That change can be registered as data."

One of the key features of the fundamental passive circuit components is that they only have two terminals but this description seems to suggest 4, 2 to each oxide layer (assuming no common point), unless I'm missing something 'fundamental'.

Flash wear-leveling is a major pain in the arse, and if this stuff doesn't need it, then even if it's slower, a lot of people will like it more than flash.

Paris, because she flashes a lot.

I grew up in the era of valves and resonant circuits, and my first thought was "Surely if this is a memory it is an analogue one?" - the "state" that can be remembered would appear to be continously variable, not binary.

That will be very interesting: an array of these would resemble the old bucket-brigade charge memory: discrete time intervals but continusly variable values.

I also don't quite get how you can interrogate them without changing them. That's a lot like the bucket brigade devices too.

Safety symbol: beware over-hype. May explode in your face.

... to see an equation explain humanity's almost total loss of common sense, especially within the voting populace.

From the article:

"In other words, it's nonvolatile memory - with a few advantages over flash. "It holds its memory longer," Williams says. "It's simpler. It's easier to make - which means it's cheaper - and it can be switched a lot faster, with less energy.""

Sounds to me like he's saying it's faster.

@Jim: You're arguing from semantics that there can be no memristor? That doesn't seem a bit daft to you?

Regards

Fake

http://www.news.com/8301-10784_3-9932054-7.html

That may shed a bit more light on the subject.

MRAM? This is the same, fast boot, remembers whats stored etc etc....and surely as it has been under development for far longer it is cheaper?

Jim, you beat me to it. Exactly what I was going to say, except I would have used more !"£$% words.

Paris, because she's !"£$%d

That sounds horribly like Heisenberg's Principle !! If you look at it, then it will be different and continue to be different each time you look at it !!

Either that or you have ingested something made from wheat that was infested with the Ergot fungus as discovered by this guy who just popped it at 102 years old !!

Icon that used to be on little squares of rice paper !!

and I don't remember anyone claiming semiconductors were a fourth base component 'element'.

Which doesn't mean they haven't.

Or have.

Why do computing clouds even fit in this dicussion? I mean, the one place you DON'T need persistant power-off memory is a 24x7 shared compute cloud, which presumably you would never want to turn off anyway. And standard DRAM seems to work just fine there.

Now, on portable, non-cloud computers, this stuff sounds interesting. Laptops with insta-on rather than sleep, mobile smartphones that don't take 3 minutes to boot, solid-state airplane blackboxes that never lose their memory, etc. A thousand and one uses, really.

But NOT in compute clouds...

But clouds are trendy and where HP wants to make money!

Does it matter if you change the value by looking at it? At least you know the original value, so you can look at the other side and change it back.

@Lee Sexton, just because something has been developed longer, it doesn't mean it will be cheaper. Mr Rolls and Mr Royce have been making cars longer than Mr Lada....

Answer from http://www.news.com/8301-10784_3-9932054-7.html

To read it you measure the voltage across it.

To write to it you apply a current.

Apparently.

(An alien because this is obviously alien inspired technology it's all a conspiracy blah blah blah yawn)

Because datacentres use huge amounts of power to run and keep cool, storage devices made from Memistor's would not need constant power and so would produce less heat as well as use less power, perfect for large scale storage. Flash memory can only be changed x number of times so this isn't really an option yet.

I think the point is, such memory would not require continuous power to retain its state. That doesn't mean the computer containing the memory has to be switched off, it just means part of the computer need not be powered up all the time. Further more, if these things end up being cheaper to produce than flash and are faster and have a higher density, maybe we'll see a viable replacement for hard-drives in a few years, saving more power, reducing heat dissipation etc. etc.

Finally, DRAM isn't that great. System architectures have to make all sorts of compromises to accommodate DRAM, so I'd say we're in need of a more flexible memory technology.

already in existence : Phase change ram.

uses a material callet chalcogenide glass as stroage element. has roughly the same speed as normal ram memory, doesn't need an erase cycle , and retains data without power.

developed by intel in 1974 ( they could not build one because of inexistent semiconductor technology. only now has semiconductor technology scaled down enough to be able to build one. ) Parts do exist. a 512k byte device is avaialble as engineering sample.

Intel developed the device in partnership with ST. ST did the production process (chemistry and manufacturing process in the waferfab ) for the production.

Its beein readied for mass production now. Samsung also has these devices ....

This new technology - and I'll be daring and say "When" it comes to market - will be unique in more than just its physical theory - it will (or SHOULD) also involve a relatively nontoxic chemistry (compared to the organic tin and silicon compounds required at present for semiconductors) in its fabrication.

Titanium dioxide is commonly used as a pigment in paints. It's not completely innocuous but hopefully it won't be anywhere near as toxic as the chemicals involved in fabricating current technology memory chips. (action: fingers crossed tightly).

"I also don't quite get how you can interrogate them without changing them."

It's called destructive read-out. Magnetic core memory, which was used until the 1970s, was based on the same principle. The only way to read it was to write it and see if you had changed it.

The instruction sets of some core-based computers actually had separate instructions for read-and-remember and read-and-forget, because restoring the previous value took some time.

And it wasn't just used in computers ..... Seeburg juke boxes from 1955 onward featured something called the Tormat Memory Unit, which takes advantage of the destructive read-out property of magnetic cores. The TMU is disposed horizontally beneath the record magazine, and has a ferrite core for each side of each record. The "letter" and "number" keys provide the half-write-in currents (with a contact in series, only made when credit is available). The destructive read-out pulse is applied to a single core (either for the A side or B side, depending on direction of travel) by means of a contact stud on the travelling player, and the amplified sense pulse triggers a solenoid which causes the playing mechanism to stop, withdraw a record, play it and return it to the magazine.

This is like a MOSFET (operating in ohmic mode) that saves its state.

@Fake.

The point I was trying to make is that a memristor device cannot be a fundamental element as it is just another type of resistance, something that is even embedded in its name. It is like saying a potentiometer is fundamental element because it can change its resistance and then 'remember' that value. But it is still, fundamentally, a resistor.

As for the existence of a 4th fundamental element, is there actually space for it?

Resistance: v proportional to i

Inductance: v proportional to di/dt

Capacitance: V proportional to int{ i dt }

What else can you do to i to require a new constant of proportionality? What Chua is effectively saying is

Memristance: di/dt proportional to int{ i dt }

Now does that really make sense?

Also, there is the mechanical anology of LCR circuits. In this analogy the three fundamental components become mass, spring and damper respectively (voltage becomes force and current becomes velocity). If there is a 4th fundamental electrical component then there must be a mechanical equivalent. Can any mechanical engineers out there provide possible candidates?

Saying all that...

The meristor design presented by HP is pretty cool as it is far, far simpler than existing flash (or any memory) cell designs.

Ok, let's go through this http://www.news.com/8301-10784_3-9932054-7.html

"The fourth circuit, which Chua called a "memristor" for memory resistor, would register how much current had passed."

Firstly, current is a measure of the flow of charge so you are actually measuring how much charge has passed.

<rant>

It seems that people really don't get electricity at all. It's like people saying that the power companies provide current. They don't, they provide voltage and you the customer draw the current you required. The best I've heard (on this site even) is people saying that the power grid provids 120v of current!

</rant>

Secondly, isn't this what a capacitor does? As charge accumulates on the plates of a capacitor (assuming your basic 2 plate model) then a potential difference is generated that is proportional to the charge. By measuring the voltage on the capacitor you know how much charge has passed, assuming that you are looking at a circuit whose initial conditions are known.

"a memristor is a variable resistor that, through its resistance, reflects its own history, Williams said."

No it doesn't. It provides an instaneous value of its resistance without any connection to previous (or future) values. If you apply a constantly varying voltage across the device you will be unable to gain any insight into the variation that occurred prior to your measurement, all you get is an average.

Other than that, nothing really in it. It is interesting that the dicussion on Wikipedia for this subject has pointed out that most material on this topic has appeared very recently and are mostly written by journos not 'experts'.

"The best I've heard (on this site even) is people saying that the power grid provids 120v of current!"

Yeah, coz any fule nose that the power grid provids 230V of current.

Yes, it is an analog process.

All computer circuit elements are analog.

The point is to drive it to the rails so that you flood out the analog to either all on or all off and then you read the "glass" as being more that half empty or more than half full.

This gets to be more painful as the circuit elements get smaller and are driven faster.

When will we see the first DSP used as a memory bus reader? (Needed because of analog distortion on the wire trace between the CPU and the memory chips. It's like the change from T-1 to DSL for Internet access.)

Someone desperate to get an 'edge' claims a new 'fundamental' wotsit so everyone gets all excited about and does the publicity for them.

From what scant information they give I'd say it's not on the same planet of fundamentality as L,C,R let alone the same ball-park

That's why, for example, memory has timings. If you read it faster than it's rated, you don't know its value.

This is also one reason CPUs tend to have clock cycles. It takes some time for the circuits involved to change state. You must wait until the state change you initiated takes place before depending on it. A clock is one of the simpler ways to do that -- just don't try anything until after the circuit has had plenty of time to stabilize.

Latches and flip-flops are needed in many circuits so that the results of computation are around long enough to read.

Digital computers are an abstraction over the analog nature of electric circuits. There's no magic involved.

"A nomenclature that suggests that rather than being a fundamental element it is derived or dependent on other fundamental elements."

Don't believe everything you hear.

than the fellow with the PHD and his think tank. You must be, otherwise how is it that you can so easily debunk this new process? And please, read the article before you spout of on this being an "element". It didn't say it was an element, it said it was a component, you know, like a capacitor, a component that holds a charge - but won't hold it when the current is removed....

I'm not electrical engineer, and I don't play one on TV, but I am sure all of the ideas you folks are throwing in attempts to trash this breakthrough have been considered or are being considered as they figure out how to bring this tech to market. I mean really, do you think HP would spend milions on something like this if they didn't think there would be a practical use for it? This is big business, not some university science project.

Paris, because even she isn't dumb enough to think that she is smarter than the folks who figured out how to do this.

Very few people think of an instantaneous resistor. The ratio of voltage to current is the resistance. This term is usually reserved for the time independent, steady state ratio of V to I.

Impedance is defined for sinusoids only in memory devices (caps and Ls). Steady state ratio of V to i is accompanied by a square root of -1.

Resistance is time independent, i.e. no time dependence at all.

Years ago I argued with a distinguished professor about the concept of instantaneous resistance, or impedance. This new component is based upon this very concept.

Note that in the IEEE spectrum article R is defined in the set of equations as dV/di.

and dq/dV=c, and dphi/dt=L, and dphi/dq=M.

These are differential relationships. You aren't used to viewing them in the instantaneous mode. Yet they exist. Thus, the relationship for M is a resistance, only it is time dependent! Just like for caps and Ls!.

It is a new component. It might be slow, it might have parasitics, it might be temperature sensitive, it might have all kinds of bugs. It is a new component folks. Not just a new kind of transistor or resistor or cap or L. This thing has a new relationship of V to i. That's big news to every charge-jockey in the world.

Wow, Computers that boot Instantly... like a C64.

Wow, Computers that will boot instantly to their previous state... So when windows crashes, Rebooting won't help anymore?

Get off your high fuckin horse. I happen to have a PhD in electrical engineering, so?

What is your problem with the term "element" rather than "component? Type "synonym element" into Google and follow the first link. Never heard of the phrase "circuit element"? Hmm, maybe if you were an electrical engineer...

I (and the majority here) never actually trashed the device just the claim that it is a new fundamental PASSIVE element/component that has never been seen before. FFS read, nothing like a good straw man eh?

C and L are time invariant, just like a resistor, as this is one of the definitions of a fundamental passive circuit element (another being that the component only has two terminals). They are not time dependent, though they are frequency dependant (mostly accompanied by lc omega or s in equations).

This new device sounds like a usefully novel voltage controlled resistor but nothing more spectacular than that. Btw, can anyone point out where the magnetic flux is in this device cos memristance is supposed to be flux/charge.

P.S. L=dphi/di not dt - but you knew that really ;-)

This thing DOES only have 2 terminals. I think this is awesome! Very useful sounding indeed, and unlike anything else out there!

To reuse a quote I found above

"The memristor consists of two titanium dioxide layers connected to wires. When a current is applied to one, the resistance of the other changes. That change can be registered as data."

(Taken from http://www.physorg.com/news128786808.html)

If you only have 2 terminals how can you apply current to one layer only and still measure the resistance of the other layer? That is truely amazing and appears to break many established physical laws. Either the device isn't 2 terminal or the description of operation (provided by HP labs) is bollocks.

Also, if the primary characteristic of the device is resistance then isn't this just a new kind of resistor, in the same way that a potentiometer or rheostat is?

- to just argue about stuff than to actually investigate it for yourself (because that's so hard on the interweb, isn't it)...

"Memristor-The missing circuit element

Chua, L.

This paper appears in: Circuits Theory, IEEE Transactions on [legacy, pre - 1988]

Publication Date: Sep 1971

Volume: 18, Issue: 5

On page(s): 507- 519

ISSN: 0098-4094

Abstract

A new two-terminal circuit element-called the memristorcharacterized by a relationship between the chargeq(t)equiv int_{-infty}^{t} i(tau) d tauand the flux-linkagevarphi(t)equiv int_{- infty}^{t} v(tau) d tauis introduced as the fourth basic circuit element. An electromagnetic field interpretation of this relationship in terms of a quasi-static expansion of Maxwell's equations is presented. Many circuit-theoretic properties of memistors are derived. It is shown that this element exhibits some peculiar behavior different from that exhibited by resistors, inductors, or capacitors. These properties lead to a number of unique applications which cannot be realized withRLCnetworks alone. Although a physical memristor device without internal power supply has not yet been discovered, operational laboratory models have been built with the help of active circuits. Experimental results are presented to demonstrate the properties and potential applications of memristors."