Intel: Metal in Future Chips = Less Leakage (updated)

securitas writes "Intel is contemplating using metal instead of silicon in future chips for the 'transistor gate, which controls whether a transistor is on or off' and the 'dielectric, an insulating layer below the gate,' which are respectively made of silicon atoms and silicon dioxide. 'Millions of minuscule switches that make up silicon chips leak electricity when they're supposed to be shut off. To compensate, engineers have increased the current, driving up power consumption, decreasing battery life for portable devices and generating more heat.' AMD has also experimented with metal instead of silicon. By moving to metal AMD and Intel expect to reduce electricity leakage. More from AP via SeattlePI and the Miami Herald." Update: 11/05 15:25 GMT by T : Read on below for some information from Intel on why this is a good thing.

gManZboy writes "Following up on the Intel news that about using metal in chips -- here's an explanation from Shekhar Borkar (Intel Research Fellow) about why heat, power, and sub-threshold leakage, not transistor size, are the real challenges to Moore's law. Apparently, in order to make chips much faster, we're going to have to pump more electricity in then anything else in our houses -- and they'll soon be as hot as a nuclear reactor -- no, really."

Comments

Which metal?

[msgmonkey]

I dont see any mention of the type of metal that would be most suitable. I'm sure all metals are n't created equal.

Re:Which metal?

[MikeD83]

In the AMD article the use of Nickel is mentioned.

Re:Which metal?

[John Courtland]

I'd also state that the writeup was a little too vague. Also, wouldn't using metal up the capacitive impedence? I'm not a full fleged EE, so I admit I could be mistaken, but I know that it's a killer when it comes to driving high frequency applications.

Re:Which metal?

[hatrisc]

they should just use platinum. that way they'll rip us off even more.... but we could sell our computers to make jewelery.... oh wait, why would any one on slashdot need jewelery?

Re:Which metal?

[Mr. Troll]

Isn't the next logical step after silicon diamond? Plenty of asian companies are already working on diamond based chips(remember, lab made diamonds are getting CHEAP and bigggeerrr)and if Intel and AMD don't get off their collective asses, they will no longer be the centers of the chip world.

Re:Which metal?

[Oo.et.oO]

the materials for a capacitor's contacts have nothing to due with it's ability to store charge (thus its capacitance). it is only a function of the device's dimensions and its dielectric.

Re:Which metal?

[dustinmarc]

It's a secret. Intel isn't saying what the material is, just that they've discovered a probable solution to fix the achilles heel of the chipmaking industry.

Basically, they are saying that they have found two new materials with a high k dielectric that reduces current leakage by more than a 100 times silicon dioxide and hopefully plan to use it by 2007 in production. They also have tested the materials and had excellent results in a lab environment. Still, they are being vague on the details, and who can blame them if this stuff is as good as they say it is.

Skeptic or beleiver? I guess that's a matter of opinion. But Intel thinks that this will keep Moore's law true for quite some time. In the past if Intel claims they have been able to stay true to Moore's law it's been true. Heck, the can't lie, Moore is the founder of the company.

Re:Which metal?

[Cougar1]

I dont see any mention of the type of metal that would be most suitable. I'm sure all metals are n't created equal.

Actually, two types of metal are probably needed. One for nmos transistors and another for pmos transistors. Nmos and pmos transistors have different threshold voltages (the voltage at which the device turns on), but ideally you would like both types of transistors to switch at the same voltage. The threshold voltage of a device can be shifted by modifying the "workfunction" of the gate metal. The workfunction is the energy required to remove an electron from the metal surface.

One reason polysilicon gates are used in conventional CMOS is that the workfunction of polysilicon can be modified by changing the level of doping and the type of dopant material (usually B, P or As). Thus, polysilicon gates can be used for both nmos and pmos transistors and by varying the doping, both types of devices can have the same threshold voltage.

Shifting the workfunctions of metals, using dopants is not so straightforward. As a result it will probably be necessary to use two different metals having different workfunctions for nmos and pmos transistors. Further complicating matters is the fact that the gate metal can interact with the dielectric material, modifying the effective workfunction and thus the threshold voltage. So, while the isolated metal may have the necessary workfunction, the workfunction may shift when the metal is part of a device. Thus, a lot of testing and experimentation is needed to find a metal that has the proper workfunction after it has been put into a device.

Re:Which metal?

[Catskul]

pardon me, "as a predacate adjective"

Re:Which metal?

[geekee]

Yes, the type of metal is important due to the workfunction of the metal. This determines the potential interactions between the semiconductor and the metal, which affects things like device threshold voltage.

Re:Which metal?

[ckaminski]

you shouldn't have posted AC. I'd WANT to take credit for that quote... ;-)

Good one.

Re:Which metal?

[Dos4ever]

Makes me wonder about the life-span of this item. Metals tend to oxidize and with current being able to carry materials, the positive end becomes doped.

Re:Which metal?

[ahdeoz]

Silicon is a metal. Your MoTB(R)(TM) implants are not going to burst and leak.

Heh

[rwiedower]

The changes are largely necessary because of the unsavory consequences of Moore's Law, the famous dictum that states that the number of transistors on a chip doubles every two years. Yeah, it's all that pesky "Moore's Law" fault...

Re:Heh

[Liselle]

They didn't even get it right, isn't it 18 months?

Re:Heh

[Alien54]

Apparently, in order to make chips much faster, we're going to have to pump more electricity in then anything else in our houses -- and they'll soon be as hot as a nuclear reactor -- no, really.

This makes sense, even from the view point on increasing density and complexity of data alone being packed into smaller and smaller containers. Even if you only allocated 1 electron per bit, after a while all of those bits start to add up. Unless you go to another system.

As an example, people often cite the human brain, with all of it's nueral connections and pathways. But this might not be all that is going on.

Biomineralization of ferrimagnetic magnetite is known to occur in a number of organisms including animals. Recent investigations have revealed the presence of biogenic magnetite in human brain tissue as well. The presence of magnetite in the brain has been established using a variety of magnetic and electron microscopic techniques.

This has interesting implications for data processing in the brain, as well a exotic areas of research into the phenomena of consciousness

Regardless of your opinion on the above (some of which is highly speculative), this leads us to the vision of a computer technology where not not only electronics states are used for data processing, but magnetic ones as well.

Re:Heh

[butane_bob2003]

Should have been called 'Moore's Prediction' cause it was little more than that. Or 'Moore's Observation'. The only unsavory consequence of Moore's Law is the imortalizing of Moore himself.

Metal Implants?

[telstar]

Man, and here I thought silicon felt weird.

Re:Metal Implants?

[betat]

No no, it'll be good. Considering that most of us are guys, silicone implants aren't very useful to us, well directly at least.

But metal..ah hah. Penis Enhancement that works! ;)

Copper?

[Anonymous Coward]

I swear I remember IBM moving to copper for chips a while back (C.2-3 years ago). Was it for production chips or just R&D purposes?

Is this just a question of Intel playing catch-up?

Re:Copper?

[msgmonkey]

You are thinking of the Copper traces instead of Aluminium, the transistors remained Silicon. Here they are talking about metal transistors.

Re:Copper?

[gus2000]

Copper is used for interconnect on the chip (and many people use it now, not just IBM). This article talks about using a metal as the gate electrode, which is these days made of heavily doped polysilicon. Copper will definitely not work for this purpose.

Re:Copper?

[DarthTaco]

You are thinking of the Copper traces instead of Aluminium, the transistors remained Silicon. Here they are talking about metal transistors.

Not true at all. The copper in IBM's process is for interconnects, not traces. I'm not sure what metal they use for the traces, but it's probably aluminum and definitely not copper. The connection between layers (interconnect) are copper plugs.

The metal intel is talking about is strictly for the gate terminal connection of the transistor. The transistor is still doped silicon or gallium arsenide or whatever semiconductor they are using.

Re:Copper?

[stevesliva]

And the entire industry has adopted copper for interconnect in modern processes, not just IBM.

Re:Copper?

[msgmonkey]

Nope, the traces are copper instead of aluminium here is an IBM article from 2000. Unless "wiring" and "traces" no longer mean the same thing.

Re:Copper?

[deadlysloth]

>>The connection between layers (interconnect) are copper plugs.

No, not exactly. The interconnect wiring on IBM chips is done in copper, however, the vias - the posts (or plugs) that connect different layers of metal in the chip - are not copper, I'm thinking they are tungsten or aluminum. I know the vias that connect the first layer of metal to the silicon transistors is a different metal than used to connect between layers of copper. Copper "plugs" tend to pop off too easy, so they use a different metal for the vias.

Re:Copper?

[Bored Huge Krill]

"traces" here is another word for "interconnect".

A CMOS chip has several layers of interconnect. In current technology, all but the lowest layer are metal (either Aluminum or Copper). That's what the whole "copper" thing was all about.

This is different; it's talking about the bottom layer of interconnect which also is used as the transistor gates. Up until now, these have almost always been some kind of polycrystalline silicon, recently with all kinds of modifications. Various manufacturers have experimented at one time or another with high melting point metals (that bit is important - you can't use Aluminum here because of processing steps that occur after the gate layer deposition) such as Tungsten or Nickel.

The use of metal for these gates is incicdental, in that the real change is in the use of a higher k dielectric. That is what makes the difference. The problem is that the polysilicon variants are chemically incompatible with the new dielectric, in the sense that the old deposition methods for growing the polysilicon layer don't work on the new dielectrics. Hence the new super-secret metal formula - the breakthrough is being able to deposit this layer on the dielectric, thus enabling the dielectric to be used - and that is what changes the leakage/Vt relationship to allow low leakage at the same time as fast switching speeds.

Krill

Silicon? Leakage?

[wickedj]

Yeah, I hate it when my silicon breaks and creates leakage.

Metal dielectric!?

[CaptainAlbert]

I have no trouble understanding a switch from poly to metal for gate connections... but a metal dielectric? That seems to run counter to common sense. The dielectric is, by definition, required to be an insulator, whereas metals, also by definition, conduct electricity rather well. What is this magic substance?

I love this site sometimes - where else can you post completely clueless questions and be virtually guaranteed to get an intelligent response from at least two people with PhDs in semiconductor physics? :-)

Re:Metal dielectric!?

[aliens]

where else can you post completely clueless questions and be virtually guaranteed to get an intelligent response from at least two people with PhDs in semiconductor physics?

Included in this great package are at least a dozen unintelligent responses coming from people who think they have PHD's in BS.

Re:Metal dielectric!?

[bpowell423]

I love this site sometimes - where else can you post completely clueless questions and be virtually guaranteed to get an intelligent response from at least two people with PhDs in semiconductor physics? :-)</blockquote>

What you meant to say was that you're virtually guaranteed to get a clueless response from at least two people claiming to have a PhD in semiconductor physics. :-) Right?

Re:Metal dielectric!?

[brassman]

A metal dielectric does sound like voodoo... but at the scale they're describing -- four ATOMS thick!? -- I suspect it's more of a waveguide (or perhaps a forcefield) than a physical barrier.

Re:Metal dielectric!?

[kinnell]

The dielectric is, by definition, required to be an insulator, whereas metals, also by definition, conduct electricity rather well

Perhaps they are using some kind of unobtanium alloy with phlogiston repelant properties

Re:Metal dielectric!?

[AdamHaun]

"With one new material -- known as a ``high k dielectric'' for its ability to hold a charge -- leakage is greatly reduced. In Intel's tests, the material -- whose composition the company would not reveal -- is thicker than silicon dioxide, but engineers say it will be shrunk."

The dielectric isn't metal, it's something else.

Re:Metal dielectric!?

[Drakin]

Well, if I recall correctly, tantalum oxide is dielectric, so it's possible that they it, rather than a pure metal.

Pure tantalum on the other hand, is a great conductor.

Re:Metal dielectric!?

[Cougar1]

The dielectric layer mentioned in some studies is Hafnium dioxide (HfO2). This is an insulator, not a conductor. HfO2, is good because it is a high-k material and it is thermodynamically stable in contact with Si.

One reason for replacing polysilicon with a metal is that the HfO2 layer is not compatible with the polysilicon deposition process. Defects form in the HfO2 layer during the polysilicon deposition step. Another reason for replacing poly with a metal is to avoid poly depletion effects. Essentially, poly still behaves as a semiconductor, so a charge depletion layer forms near the poly/dielectric interface. This depletion layer acts as an insulator and has the same effect as increasing the thickness of the dielectric layer (which is what we're trying to reduce). The increased thickness reduces the capacitance, which needs to be large for the transistor to function properly. Unlike semiconductors, metals do not form a depletion layer.

Re:Metal dielectric!?

[inl101]

I'm about 6 months away from my PhD in semiconductor physics.

They mean metal oxides. Leading candidates are Halfnia and Zirconia. These are "High-K dielectrics".

Using both reduces the Effective Oxide Thickness (EOT) of the gate dielectric. For the same thickness material, high-k dielectrics look like a thinner amount of silicon dioxide. Metal gates eliminate depletion effects in the gate (poly-depletion), which also makes the oxide look thinner.

With lower EOT, the gate has better control of the channel, so leakage goes down.

Re:Metal dielectric!?

[Cougar1]

Tantalum oxide is a good high-k dielectric, but it is not thermodynamically stable in contact with Si. As a result, Ta2O5 reacts with Si during the high temperature (>900 C) anneals necessary to activate the Si dopants. These unfavorable reactions ruin the devices and as a result Ta2O5 has largely been abandoned as a potential dielectric in Si transistors. Ta2O5 is used for capacitors in DRAM memory devices.

Re:Metal dielectric!?

[stevesliva]

Ta2O5? Does that go something like, "Ta-Ta! ooooo"

Definitely slashdot appropriate material.

Re:Metal dielectric!?

[EuropeanSwallow]

Not necessarily for the case in order, but metals are not by definition good conductors. Generally metals are good conductors for a given frequency range.

Even at 50-60Hz and in the case of power conductors, harmonics suffer what is called pelicular effect, which basicaly is the fact that these harmonics, as their frequency rises, propagate gradually towards the outer layers of the cable, raising the effective resistence of the cable "seen" by that harmonic (less area to conduct->resistance proportional to the inverse of the conductive cross-section).

If the frequency should become high enough, it would become in the end non-conductive. Thats why you need waveguides (at those frequencies air is the good conductor!). Thats also why printed circuits for very high frequency are extremely bulky.

Re:Metal dielectric!?

[Midnight Thunder]

The important thing to take into account is that a transistor is essentially a non-mechanical switch. Anything that can achieve the same result, semiconducting or not, could be used. At the same time when we are at the scales that are involved in a microchip then it is also important to have molecules that will not react with molecules making up other aspects of the chip.

Like many subjects, if you are basing your knowledge on what you find in main stream text books then you are going to be missing a lot. The best sources are journals, since they tend to have the latest research.

Like many on /. I lack a good background to give a good explanation of what is going on, but a quick search on the web for 'metal-dialectric', shows that there is certainly some research being done in this field (no pun intended).

Re:Metal dielectric!?

[Fulcrum of Evil]

The dielectric isn't metal, it's something else

<astrophysicist>So, it's He?</astrophysicist>

Re:Metal dielectric!?

[gantrep]

yeah

Re:Metal dielectric!?

[Cougar1]

and the reason it has to be in contact with Si is .... ?
remember they're talking about replacing silicon altogether. a Ta2O5/Ni (or whatever) transistor with copper over Al interconnects and no silicon whatsoever would work fine.

They are not replacing all of the silicon with metal only the polysilicon gate. The underlying material must still be a semiconductor and in this case Si. The High-K dielectric layer sits between the gate and the underlying Si. Thus, replacing polysilicon in the gate with metal solves the problem of reaction at the top interface, but the high-k material is still in contact with silicon at the bottom interface.

Re:Metal dielectric!?

[LadyLucky]

I mightn't be able to help you out, but I think that Britney Spears knows a thing or two.

Re:Metal dielectric!?

[jaoswald]

I believe what you are speaking of is more commonly known in English as the skin effect. The frequencies for metals to become non-conductive are quite high, as can be demonstrated by looking in a mirror. The reflection you see is due to visible electromagnetic waves (over 10^14 Hz) causing conductive currents in the metal. Metals are still conductive until you reach the ultraviolet, when you exceed the plasma frequencies, so that electrons can no longer keep up with the field oscillations. At that point, just about everything begins to look like a lossy dielectric, because everything is made of atoms.

For microwave-frequency waveguides, the air is not conducting ordinary current. ["Displacement" current is another matter: changing electric fields contribute to Maxwell's equations similarly to flows of electric current.]

What matters about air is that it is low LOSS. Even good conductors have some loss in them, so transmitting charge currents causes ohmic losses and inefficiency. Transmitting electromagnetic waves in dielectrics [=NON-conductors] reduces the loss. You still need good conductors to create the guide walls, until you get to optical frequencies and use different refractive indices.

Even at 60 Hz, the electric energy going through lines is in the fields between the lines, not within the wires. The losses are due mostly to charge currents flowing within the highly conductive wires.

Re:Metal dielectric!?

[JGski]

The issue with both copper and Hi-/Lo-K dielectrics is that they are chemically incompatible with silicon and many other tradional materials, that is, if they are allowed direct physical contact, either the silicon or the dielectric is irreparably damaged. For example Si sucks up Cu like a sponge and ceases to be a semiconductor in the process.

For this reason you add barrier metals between Cu & Si, or barrier dielectric jackets between Cu and Hi-/Lo-K, or Poly-Si and Hi-/Lo-K. These are often are weird-*ss transition metal compounds: Hafnium Silicide, Scandium Silicide, Tantalum Silicide/Salicide or complex dielectric compounds or structures. It's like a periodic table reunion.

Figuring out what works requires building a lot of sample devices, and then doing electrical reliability testing on them followed by failure analysis. Very expensive and time consuming, thus perhaps not something you want to share with free-riding competitors.

Re:Metal dielectric!?

[John Courtland]

Does it react the same way with GaAs?

Re:Metal dielectric!?

[Cougar1]

I'm not sure whether Ta2O5 would be thermodynamically stable in contact with GaAs. However, it probably wouldn't be an issue due to the much lower process temperatures used for GaAs. However, even if thermal stability is not a problem, there is no guarantee that the Ta2O5/GaAs interface would be of sufficiently high quality. Defects at this interface cause fermi-level pinning, degrading device performance. There is some promising work for Ga2O/GaAS reported here that might be of interest to you.

Re:Metal dielectric!?

[Kokeshi Harinezumi]

All this talk about dialectrics is making me dehydrated.

Re:Metal dielectric!?

[yomegaman]

So, you're saying I should pay out the ass for Monster Cable for my speakers, then? Wow, the guy at Best Buy was right after all! ;-)

Re:Metal dielectric!?

[doug363]

I've got another question, and this seems to be a good place to post it:

I understand that gates were originally metal in the early days of FETs, but then were changed to polysilicon. IIRC, one of the reasons was that metal gates are much more susceptible to electrostatic discharge (this only affects interface transistors, though, not the internal ones). And I seem to recall that polysilicon can generally be lithographically etched thinner than metal can, so you can get shorter transistors and smaller gate capacitance. Is the idea of a metal gate that revolutionary? The article at news.com.com suggests that the metal gate stops phonon scattering.

Am I correct about the early use of metal gates? And how does the metal gate absorb phonons or prevent scattering? news.com.com says something about reducing electron mobility (I assume this wouldn't affect electron mobility in the channel... not something you really want to decrease).

Time for change...

[stm2]

to MetalValley!

Now, instead of "experiment in silico", it would be "in metal" (??) or "in Fe|Au|Cu" :)

Re:Time for change...

[Seahawk]

Ah well - with the downturn in the IT industry, we could just rename Silicon Valley to Nickel Valley... ;o)

Re:Time for change...

[hal9000]

Wouldn't work. A place called MetalValley would attract a bunch of washed up drunk metalheads looking for shows, and it'd be like high school all over again for the poor geeks.

On the other hand. I'd definitely go there to see shows. :-)

What about...

[the_bahua]

...diamonds?

I thought that the manufacture of diamonds was set, and only needed to step up its production. Gemesis has been making, for less than $100, gems that would be worth hundreds of thousands if naturally mined.

The most promising thing about these diamonds is that, being cheap, they open the door for cpu cooling. Diamonds are tolerant of exponentially higher temperatures than silicon, so why aren't we hearing about intel, amd, motorola, ibm, TI, and sgi taking advantage of this new technology.

Metal? What about metal is unprecedented? What about it has kept us from using it before? Diamonds are the future, not metal.

Re:What about...

[October_30th]

Intense lobbying, FUD and outright threats from the diamond industry have managed to suppress any large scale production of perfect diamonds (you can't do chips using crude industrial grade diamonds).

You see, diamonds are seriously overpriced luxury items. Although it is possible to manufacture cheap diamonds that are indistinguishable from the natural ones, it has never been done. Why? It would ruin the entire business model of De Beers & co. which is based on artificial scarcity. That's why they'd fight such projects to the bitter end.

Re:What about...

[tyroney]

Did you miss that whole bit about how there are at least two people that are making rocks that are only distinguishable from "real" diamonds because they are better?

Just checking, because it souded like these people were hoping to do just what you mentioned as being heavily fought. And so far, they haven't been killed as far as I know.

I, for one, welcome the death of our diamond-scarcity-based overlords.

Re:What about...

[l3prador]

Actually, it's currently being done by Apollo Diamonds and Gemesis, which was mentioned above. De Beers is fighting them as hard as they can, but even if they convince the public that manmade diamonds aren't worth anything as jewelry, they will still be able to use them for computing. However, production is not quite ready for large-scale chip manufacturing, which is why Intel and others have not yet turned to diamonds.

Re:What about...

[October_30th]

They don't have enough visibility or scale, yet.

Re:What about...

[Chazmyrr]

And so far, they haven't been killed as far as I know.

Yet. Even if they aren't physically killed, the cost of defending all the various lawsuits and/or lobbying for legislation by De Beers may just kill the companies.

Re:What about...

[mentaldrano]

Diamonds? Until the diamond fabrication process becomes much more advanced, diamonds are a waste of time. Impurities are the culprit here. Many impurities = low mobility electrons = crappy chips.

Electrical grade silicon (EGS) has a long purification process that it must go through to be of sufficient quality to make chips from. To give an example, there are roughly Avogadro's number of silicon atoms in one cubic centimeter of silicon (5.5x10^22 atoms / cc). After being purified, the MAXIMUM impurity concentration is ~10^14 atoms / cc. This is around 1 part per 100 million. The best laboratory grade Si is 100 times better than that!

Contrast this with diamond: even the best artificial diamonds have so many impurities that you can see them with the naked eye. So even if diamond semiconductor chips can be made (SiC is much more likely) we won't be able to use them for anything like what we do now until the purification process is improved.

Re:What about...

[fermion]

I think I have mentioned this before.

if you are talking about diamond films, the technology is available and probably just needs to be scaled to production. I think the research is there to dope as needed. I don't know how useful diamond is as circuit material.

If you are talking about bulk diamond, there is a large difference between creating a diamond for jewelry or the drilling industry and creating a diamond boule that will produce 8+ inch wafer for manufacturing. Beyond coming up with a reliable source of seeds that will actually grow a boule big enough to produce sufficiently large wafers in sufficient quantities, the cutting process is likely to be more expensive than silicon.

Re:What about...

[Slime-dogg]

I just love military people. You get the most flavorful language out of them, as well as amusement:

So, for now, Clarke is sticking with cultured. But in the end, he insists, it won't really matter. "If you give a woman a choice between a 2-carat stone and a 1-carat stone and everything else is the same, including the price, what's she gonna choose?" he demands. "Does she care if it's synthetic or not? Is anybody at a party going to walk up to her and ask, 'Is that synthetic?' There's no way in hell. So I'll bite your ass if she chooses the smaller one."

Such bluntness and resistance to limp-wristed caution is an attribute to be emulated and admired.

Re:What about...

[Prune]

That is simply not true. Gemesis and Apollo Diamonds have produced samples purer than the best natural diamonds.

Re:Alien Technology?

[SharpFang]

Now how can you say that CPUs are based off of alien technology when Intel is making changes like this?

They just caught a new flying saucer.

I'm no expert, but...

[doublebackslash]

I suspect that the meatal they are going to be using will be doped with something else, or impure or part of an oughtright compund. Just like "Sodium" in your diet just happens to be bonded to some chlorine. I have no idea how they get away with this sort of thing.

Diamond

[rogue555]

Just make 'em out of diamond and all your problems are solved. Great heat transfer coeficient so you don't have to worry about heat anyway.

From Intel's site...

[sczimme]

The history of Moore's Law.

Or if you are interested in Moore's original paper, you can find it here.

Progress!

[keoghp]

I see they're still trying to push the envelope. Perhaps what they should be doing in folding it over, sticking it down, and posting it.

Clueless, thy name is reporter

[overshoot]

To clarify: the idea is to use a gate dielectric which has a higher dielectric constant than silicon dioxide. Most of the candidates are metallic oxides, nitrides, etc. That allows the transistors to have thicker gates for the same gate capacitance (which is how MOS transistors work).

The chemistry of the non-silica gate dielectric requires that the gate itself be non-silicon, and metals are better conductors anyway. (For larger transistors, we're already running into trouble from the distributed resistance of the gates.)

Hope that helps.

Re:Clueless, thy name is reporter

[Anonymous Coward]

The article make no mention of what the gate is composed of, but they offer a hint. The confusion comes from calling it a "metal gate". Back in olden days, we only used metal gates composed of Al. Unfortunately, there was a limit to how thin the silicon dioxide gate insulator could be made, because the Al would diffuse through and kill the transistor. This is why polycrystalline silicon (poly-si) is used for the gate conductor material.

There seem to be two advances proposed here. One is to replace the poly-si gate conductor with a metal conductor of unknown composistion. Al is likely because of its current use in lower level interconnects (Cu can't be too close to the Si). Al has way higher conductivity than poly-si, so there would be a speed increase.

The "high-k" (high dielectric constant) material is a metal oxide, likely TiO2, ZrO2, or HfO2. These are dense films, which would act as a diffusion barrier for the Al gate, circumventing the old problem solved by the poly-si gates. In addition, the higher dielectric constant (20-50, or so, compared with 3.9 for SiO2) would greatly increase gate capacitance, which directly affect transistor gain (increased speed).

Re:Clueless, thy name is reporter

[Cougar1]

Al isn't really practical as a gate material, since the melting point is too low and it wouldn't survive the high-temperature anneals necessary to activate the dopants in the underlying silicon. Instead the gate material will be some kind of refractory metal, like Ti or TiN or something similar.

Also, another reason for using polysilicon is that you can shift its workfunction by changing the doping from p-type to n-type. Thus, polysilicon can be used for both nmos and pmos transistors.

Replacing poly with a metal will require either using two metals with different workfunctions, one for nmos and the other for pmos, or using a "mid-gap" metal with an intermediate workfunction and doing some other tricks to compensate for the differences in threshold voltages of nmos and pmos devices. One of the key challenges of replacing poly with metals is finding metals with the correct workfunctions.

Long Run 2

[glenrm]

Seems like a lot of work when the could just license Long Run 2 technology from Transmeta.

Re:Long Run 2

[gorilla]

Obviously the idea is to make Intel's own implementation, so they can then go and sue Transmeta for infringing Intel's patents.

Re:Long Run 2

[glenrm]

If only that was funny and not a possibility... The implementations do seem to be quite different one is software controlled and available today (LongRun2) and the other is material based and will be available at some time in the future...

Re:Long Run 2

[butane_bob2003]

Exactly. Intel has been pushing the 'how many transistors can we fit on a chip' envelope for a while. I wonder if they are exploring the 'how much can we do with fewer transistors?' question.

Re:Long Run 2

[Kokeshi Harinezumi]

> Seems like a lot of work when the could just
> license Long Run 2 technology from Transmeta.

Would not the worst-case scenario of Long Run 2 technology be near zero effectiveness?

I don't suppose to know the circumstances that might lead to this scenario, but if Long Run 2 cannot guarantee that it will perform sufficiently in *all* cases, then there must exist an instance where a processor depending on it exclusively would overheat.

Thus it would be an ineffective solution, as opposed to using a physical material that will *guarantee* itself by virtue of its chemical properties.

Re:Long Run 2

[glenrm]

Overheat? Isn't it meant to address leakage not just thermal issues? Your logic also is flawed in that you imagine a worst case scenario that you can not describe and then due to this scenario choose a non-exsistant solution over one that exsist today.

Isn't this old news?

[wazzzup]

I thought it was already established that silicon implants were prone to leakage.

But switching to metal? Man, I'd hate to walk outside on a cold Montana morning in February with those.

What's that? Silicone? They're not the same? Never mind. Carry on. Sorry.

So you're telling me SOI is NOT a busty gal in an angora sweater?

Re:Still Binary...

[Valar]

Of course, i286 instructions only apply to i286 sucessors. There are lots of companies that that have actual innovative architectures (or at least aren't afraid of a little loss of backwards compatibility). Quantum computers are, of course, a much different thing. The "instructions" for a quantum computer are about how to decohere data without lossing it, as oppose to today's instructions which are "move these numbers, do an operation, move them again, etc, etc." They will be useful for computation (replacing what is currently done by brute force number crunching right now), but they don't serve any particular use on the desktop (you mentioned the i286, so I can only assume you are talking about desktop centric machines). So no quantum Halflife 8 for you. nyah.

Hadn't IBM already done this

[adzoox]

I was aware that IBM's copper on silicon insulator already acheived less leakage and less power consumption, also increasing power (per Mhz) in each cycle. G3's (for Apple Computers have had this for over 2 years) and G5's also have it.

Interesting how IBM has discovered that moving to metal for processors and away from metal for hard drives. (Newest Hitachi/IBM notebook drives use Pixie dust which is actually glass. The platters in these hard drives are also ferro impregnated glass platters)

Re:Hadn't IBM already done this

[Oo.et.oO]

you are thinking of SOI (silicon on insulator) which allows for less DRAIN current leakage to the substrate. this of course has nothing to due with the copper interconnects in the BEOL.

all existing technologies in production (AFAIK) use poly gates as it survives the anneal and etching steps which copper and aluminum could never do in current configurations

Re:Hadn't IBM already done this

[adzoox]

No, the SOI is something different - it IS used in the G5's but G3's (750fx & Gx) used copper interconnects as well. It was the way that IBM figured out how to make the G3 so effiecient. The 900Mhz G3 is probably the coolest/best performing/per Mhz of any processor released in the past 3 years.

Re:Hadn't IBM already done this

[Remlik]

Yea, HD producers have been using glass platters since the very first notebooks came out.

I remember dropping and breaking 2 gig Arial Hard Drive on the concrete floor of the tech room I worked in around 1994. Took the top off the drive and it was all crushed glass.

But I'm sure IBM came up with it anyway.

Re:Hadn't IBM already done this

[mentaldrano]

No, IBM has gone to copper interconnects, which have lower resistance than the current aluminum ones that everyone else uses. IBM's innovation was finding a way to keep the copper from sinking into the silicon and ruining the delicate transistors underneath.

Intel is actually talking about replacing the gate dielectric (which is silicon dioxide currently, even at IBM) with a metal or metal oxide, which has a higher dielectric constant. Higher dielectric constants mean a more effective gate for the same thickness, or the same gate effect for a thicker layer (and hence less leakage).

Intel is also apparently talking about replacing the polysilicon gate with an actual metal gate. Polysilicon is used for gates because it doesn't melt when the chip is annealed (an important processing step), like metal would. Intel's innovation is apparently figuring out a way to get around this problem.

Re:Hadn't IBM already done this

[Anonymous Coward]

That's not quite right. IBMs process uses copper as an interconnect between transistors. The contact to the device itself is still poly-Si.

when hooking devices together it goes something like:
transistor -> poly-Si contact -> TiN buffer -> Cu interconnect

There are 2 problems with putting a metal directly on Si:

1) electronically they don't match up and you get a barrier to conduction. This means you have to apply a LARGER voltage than you would with Poly-Si to even get it to conduct at all. The band structure of Si matches up with Poly-Si pretty well so the barrier is small, but the trade off is that Si is not as good a conductor as the metal.

2) Metals often react with Si and kill devices. Copper is especially good at this, which is why Aluminum was used as an intereconnect before IBM figured out how to do it right. The answer was a buffer between the Si and the copper that was a pretty good conductor but didn't react with either. TiN is common for this purpose.

Re:Hadn't IBM already done this

[addaon]

The 900Mhz G3 is probably the coolest/best performing/per Mhz

Except of course for the same chip (the 750FX) at, say, 600MHz, or less. The G3 is seriously bandwidth-starved in most configurations I've seen (it supports a 200MHz FSB, but I've never seen it used with more than 167); scaling down the clock-speed gives sub-linear decrease in performance, linear (well, close enough; moreso than for most non-arm chips) decrease in heat and power consumption.

Don't get me wrong, the 750FX is, in my opinion, the nicest piece of silicon yet produced (and I'd much rather have a 4-core G3 than a 1-core G5 with the same number of transistors)... but people (and, of course, in particular apple) need to realize that this chip had dual phase-locked-loops and takes less time to switch clock speeds than to switch processes... dynamic scaling is Good, constant 900MHz (or 900MHz and then, blindly, 600Mhz when unplugged from the wall) is Silly and Bad.

Re:Hadn't IBM already done this

[RzUpAnmsCwrds]

AMD's Athlon 64 can change frequency and voltage on the fly to any of several levels. Even the desktop version has this feature.

Intel has been using copper interconnects since Pentium III "Coppermine"; AMD since Athlon XP "Palomino". AMD is currently (AFAIK) the only company producing CPUs using a SOI process.

Re:Hadn't IBM already done this

[haxor.dk]

Well, CPUs and hard drives are two different beasts entirely.

One deals with flow of eletricity (erm... well... voltage switching, really), the other with stable magnetization.

Re:say what?

[stevel]

Why is VLIW not more popular? Because compiler technology isn't yet good enough and current VLIW designs have restrictions that get in the way of the best performance.

Over the years, there have been many attempts to use techniques such as VLIW, which sound great on paper, but don't do well in practice. What have worked the best, at least through the 90s, are architectures that do a lot of simple things fast.

You can make VLIW fast, Intel has managed that, but at great cost in both silicon and software.

Be careful when making generalizations about a processor line such as the P4 - there have been quite a few P4 generations, each better than the last. Latencies have gone down.

I think that parallelism (eg. HyperThreading, multicore, etc.) is where the real-world performance gains will come from. Single-threaded benchmarks don't accurately reflect realistic workloads.

Re:Still Binary...

[SharpFang]

but they don't serve any particular use on the desktop ...yet.

Think about computers in their "elder days". They were useless for desktops too.
So I certainly won't see HalfLife 5 on one, but 8... why not? It's just matter of hardware, combining it with current technology (think "hybrid computers", where most of the stuff is done "binary way" but computations that would take hours on P4 are solved within time of, say, 8 "classic CPU cycles" (of which most would be uploading and downloading input and output anyway), by the quantum FPU.

*Switching* to metal?

[Hydrogenoid]

Isn't Silicon a metal?

Re:*Switching* to metal?

[overshoot]

Isn't Silicon a metal?

No. The bonds between silicon atoms are covalent. A metal (e.g. copper) has a "cloud" of electrons free to move around in the lattice. Silicon is a semiconductor, with the charges bound to the atoms except when there's enough energy (typically thermal) to kick them loose.

Re:*Switching* to metal?

[crgrace]

No, silicon is a semiconductor, which means that its valence and conduction bands are not overlapping (as in a metal) at room temperature. It's not an insulator because the bandgap (the electrical distance between the valence and conduction bands) is small (about 1.1 eV).

Re:*Switching* to metal?

[brokenbeaker]

of course, the trick to industrial use of Silicon in making chips etc. is the proper doping of the silicon. subsituting few ppm Si with an element that has fever valence eletrons produces an excess of holes, which conduct electricity (p-doped Si). the opposite makes for electron rich (n-doped) Si. the conductivities of these materials is much higher than pure Si.

Re:say what?

[October_30th]

One thing that perplexes me is why VLIW is not more popular.

x86 is here to stay because people want to run legacy code reliably and efficiently. Intel tried to get rid of x86 architecture with Itanium, but just watch how they'll eventually have to release a version with the crosslicensed AMD's (horror of horrors!) x86-64 instruction set.

They call it Low-K Dielectric

[OS24Ever]

Here is an article explaining low-k dielectric. I believe this is a shipping product on the Power4/4+ based systems and it is in the EXA chipset on the x365/x440/x445/x450 Intel servers, and the Apple G3 and G5. The xSeries products even have little copper BB's in the grill of the system to symbolize that they use copper based technology.

Re:They call it Low-K Dielectric

[ggram]

The copper isn't the low-k dielectric, the dielectric is what is in between the copper wires, it is what determines capacitance (and actually is should be called high-k). Using copper over aluminium reduces the resistance of the wire, but using a low-k dielectric lowers the capacitance, both of which make things faster.

Re:They call it Low-K Dielectric

[overshoot]

Different subject. Up in the metal interconnect layers, you want low-K to reduce capacitance between unrelated signals. Down at the gate level, you want high-K dielectrics to make it possible to induce a reasonable channel charge at low voltages and practical gate thicknesses.

This might help save the world (for a little while

[basingwerk]

This invention might save energy and lower carbon emissions. This is a good thing if it helps delay the next ice age, which is due in a couple of decades when the gulf stream conveyor collapses due to global warming. No use making very long-term plans, but anything that can delay the freeze until I move on to the next world must be tried!

Gee, moving back to metal gate fabrication?

[dido]

My course in VLSI design was many, many years in the past, but what I do remember is that early integrated circuits used metal gates in the fabrication process. That process was later abandoned in favor of polysilicon because poly was much easier to work with at smaller feature sizes (I'm a bit foggy on this one). Gee, so now we're going back to metal gate processes, and we'll have real metal-oxide-semiconductor field effect transistors again?

If this is becoming easier to do at deep submicron level, I suppose processes for making deep submicron feature-sized Gallium-Arsenide MESFET's also got easier? Now wouldn't we just love to have such GaAs chips on our desktops... (I do know I'm forgetting another difficulty in working with GaAs, anyone care to remind me why GaAs is not as common as silicon today?)

Re:Gee, moving back to metal gate fabrication?

[overshoot]

That process was later abandoned in favor of polysilicon because poly was much easier to work with at smaller feature sizes (I'm a bit foggy on this one).

Silicon gates can be self-aligning. Once you've got gate oxide, deposit a layer of polysilicon and pattern it, then use the remaining poly as a mask for the gates while the rest of the oxide is removed.

I do know I'm forgetting another difficulty in working with GaAs, anyone care to remind me why GaAs is not as common as silicon today?

There are several. The defect densities in compound semiconductors are much higher than silicon, limiting size. The materials aren't as mechanically stable as silicon, which also limits size due to misalignment. Also, complementary FET structures are hard to get right (which is why most compound semiconductor circuits today are basically bipolar.)

I suspect that I'm also forgetting a few. It's been a while for me, too.

Re:Gee, moving back to metal gate fabrication?

[Shiifty]

If this is becoming easier to do at deep submicron level, I suppose processes for making deep submicron feature-sized Gallium-Arsenide MESFET's also got easier? Now wouldn't we just love to have such GaAs chips on our desktops... (I do know I'm forgetting another difficulty in working with GaAs, anyone care to remind me why GaAs is not as common as silicon today?)

Its been awhile since I've worked with semiconductors, but GaAs has a direct bandgap and Silicon has an indirect bandgap. I think GaAs is more suitable to lasers and other opto-electronic devices, as indirect bandgap materials are inefficient at emitting light.

However, electronic devices can be made with both direct and indirect bandgap materials. GaAs is a costly material with high processing costs compared to Silicon, and hence not suitable for mass production (at this time).

Re:Gee, moving back to metal gate fabrication?

[empty]

...anyone care to remind me why GaAs is not as common as silicon today?

GaAs does not have a robust oxide. The entire silicon industry expanded because silicon is easy to oxidize and provides an excellent dielectric. If this were not the case, we would probably still be using germanium, because it is easier to purify (and/or GaAs, because it is faster, higher temperature, and direct band-gap).

The funny thing is...

[philthedrill]

gates used to be metal in the first place (before polysilicon). MOS = Metal Oxide Semiconductor.

Re:say what?

[tomstdenis]

Why not do something like the Transmeta processors though? Even in a desktop I can't imagine that being a bad idea.

E.g. write CMS programs for x86, PPC and heck allow native VLIW too.

If the Efficeon gets the performance per Watt they are claiming that's amazing. At 7 watts that's about 10 times less than the average P4/Athlon desktop processor.

Tom

Global Savings?

[SmashPot]

With the massive number of computers today, wouldn't this have a dramatic affect on the total power consumption at any given moment? I mean, initially the impact will be less, but as the world replaces their PCs with new chips using metal switches, I would think this would have a substantial effect on the power grid. Additionally one would think this might also increase the life of the processor. They do not mention if this will cause an increase in the cost of manufacturing though. I would assume so since silicon is less expensive than metals

awesome

[MoFoQ]

lower the power consumption equals less heat which in turn means they can push the envelope a bit more and voila! Fast cpus. Just when ppl start to doubt moore's law...they get a swift kick in the groin called technology.

Radiation Hardness, does this do anything for it?

[JonnyRo88]

It would be awesome if this move would improve the electromagnetic pulse response characteristics of the chips that use it. Regarding past slashdot articles on EM pulse weapons, I would like to see even consumer technology become more pulse tolerant.

I cant help but worry that EM pulse weapons designed to take out electronics in a small area could become more commmon, used by either terrorists/pranksters. I'm not talking about it happening today, but in the future maybe 10 or 20 years down the road, I think it highly likely that there will be a publicized case of someone using an EM pulse device to take out a business computer network.

Why is it that when people talk about EM pulses taking out electronics that no one thinks about people who might be wearing pacemakers. I would hope that those devices are fault tolerant enough to save the wearer in a large EM pulse situation. Perhaps the body supplies some shielding, but I wouldnt know. Lots of people these days are running devices that are essential to life.

The metal is Nickel... for AMD at least

[fruey]

Switching to metal gates and high-k gate dielectrics also eliminates phonon scattering. Increasingly, the atoms inside transistors are vibrating. Incorporating high-k gate dielectrics alone does not solve the problem.

"This slows down the mobility of electrons," David said. "The metal gate seems to act like a sink for this phenomenon."

David, however, declined to identify what metals Intel is experimenting with. AMD is working with nickel in its metal gates.

So it's all a bit over my head, nickel certainly conducts electricity but obviously not like gold and at an atomic level maybe it doesn't matter... since the silicon is like 4 or 5 atoms thick, and the metal seems to be some sort of a sink for lost electrons.

Physics PhD to the rescue?

Re:The metal is Nickel... for AMD at least

[overshoot]

Nickel is the metal on top of the gate dielectric. Most of these processes actually have layers of material in each structure, though. For instance, the dielectric may have a layer of something compatible with silicon at the "bottom" and then something with high K above it, then something on top of that which gets along with the metal of the gate itself. Nickel is good because it's fairly nonreactive with the other materials in use.

Re:The metal is Nickel... for AMD at least

[mentaldrano]

Yes, metals (like Nickel) do conduct electricity, but most metal oxides do not! I suspect the mystery Intel metal and the AMD Nickel dielectrics are actually metal oxides. The benefit of metal oxides is that they have huge dielectric constants (which make better capacitative gates). I believe the world champion dielectric material is actually an oxide of Tantalum, but I cannot remember off the top of my head.

As for the phonon question: in crystals the quantum of atomic motion is called a phonon. Electrons can scatter off of phonons, reducing the mobility and hence increasing the resistance. Two ways around this: lower the temperature, which suppresses the creation of phonons, or use a heavier material, which is harder to move and hence phonons take more energy to create. Using a metal gate dielectric (heavier material) traps any phonon which touches it, reducing the concentration of phonons near the surface, where the conduction electrons are.

Re:The metal is Nickel... for AMD at least

[JGski]

The nickel is probably a barrier layer to prevent chemical mixing of Cu or HiK with other materials in the device. Usually barrier layers are real thin, but are absolutely essential to prevent the device from morphing spontaneously into a glob of uselessness

99% of all conduction in semiconductors is "phonon"-mediated/impeded. Metals conduct through a different mechanism. I doubt metal gates significantly affect semiconductor conduction through phonon modulation, especially since there is no direct contact with the active gate region and metal.

Some explanations and thoughts

[fredrikj]

Would metal really be able to replace silicon? IANAEE, but...

Wait, that only works on the law forums. Darn.

Re:Some explanations and thoughts

[Mizery De Aria]

I am not an ee? ee being?

Heat=power

[nagora]

If the chip in your computer is as hot as a nuclear power station, should you not do what power stations do and hook it up to a steam turbine?

One day, your computer may be the ONLY thing in your house connected to the outside mains supply!

TWW

Re:Heat=power

[levin]

That would work great except for two things:

(1) Power is really related to heat and volume, not heat alone. Ie, it may be as hot at a nuclear reactor, but it can't heat the same volume of water as a nuclear reactor can to boiling.

(2) Processors are not designed to generate heat, this is just a byproduct, so the Power consumed by the chip is probably MUCH greater than the power delivered by the chip as heat. Since you'd have to pump an insane amount of power into the thing to get a little power out, you would not be able to run your house economically.

Re:Heat=power

[joto]

Power is really related to heat and volume, not heat alone. Ie, it may be as hot at a nuclear reactor, but it can't heat the same volume of water as a nuclear reactor can to boiling.

This is nitpicking. The chip vs nuclear reactor heat example came up by extrapolating how todays CPU's use power, and by assuming that we would use similar ways to achieve better performance in future CPU. You can extrapolate this as long as you want. Since the power grows exponentially, soon enough you will have a sun or a supernova. But most people would probably realize that something requiring more than a few nuclear power plants to power it would be unpractical. In fact, most people would probably reject owning a computer that consumed just a few kilowatts.

Processors are not designed to generate heat, this is just a byproduct, so the Power consumed by the chip is probably MUCH greater than the power delivered by the chip as heat.

I would then like to ask you a simple question: Where does the rest of the energy end up at? If it's not converted to heat, it needs to be converted to some other form of energy, such as kinetic energy, light, etc.. (which in turn will sooner or later end up at heat, but possible elsewhere). Since a CPU doesn't do much physically interesting stuff in meatspace, I think it's safe to say that they are very effective heaters.

Hmm...

[Bendebecker]

Does this mean I should start investing in liquid nitrogen suppliers? I hear pools develope leaks all the time. What will happen when the liquid nitrogen starts to drip? Note to self: don't store computer over anything valuable.

Re:Hmm...

[HermanZA]

Liquid Nitrogen drips? Well, it will just turn into gaseous Nitrogen, which is commonly known as air - no sweat...

Re:Still Binary...

[VanWEric]

With a few more source code leakages, I am sure we can get 5 to be pushed back to the point where it is quantum.

pursue the potential pittsburgh property pronto!

[fluxrad]

nt.

If the dielectric gets thinner,

[awfar]

...the physical stress on dielectric becomes much greater (the square of the distance?) and more prone to breakdown and breakthrough; though, the research is all about eliminating these problems in a practical sense. As you get closer(thinner/lower-k), leakage increases, so a balance will be made once again between leakage, feature size and geometry, materials, thickness of oxide, and voltages.

Wow, I'm speaking Japanese! And, I have never heard Japanese before, so I'm not sure what I said.

Re:say what?

[Zathrus]

Has anyone actually checked the specs of the P4? Things like 15 cycle multiplies, 1.5 cycle ADC/SBB, etc

And yet it's still faster than virtually any other processor on the planet. Intel has come a long way from when it was being spanked by MIPS/Sun/etc. You can make an argument for Alpha, but that's about it.

a non-x86 core

The P3, P4, and Athlon cores aren't x86. They have a wrapper layer that translates x86 instructions into their own internal core instructions, but that's it. And, frankly, a more "efficient" core doesn't make a bit of difference if it doesn't actually have any use in the real world. The x86 ISA is here to stay for a long, long time. People have been predicting it's death since it came about, and yet it's managed to dominate every other ISA out there. Hell, it's being looked at for embedded use now of all things.

Why should the CPU do the work of a compiler at runtime?

Because the compiler doesn't know what the dataset is. You can make guesses, but that's all. If you really want to optimize then you have to actually run the program for a period of time using real world data and then re-compile with the profiling data you've gathered. Which is pretty damned expensive to do, and is invalidated if your data set changes (yeah, that never happens in the real world) or you want to sell the program to multiple companies (again, one of those rare edge cases). The fact of the matter is that it's far, far cheaper to upgrade the hardware than it is to spend a bunch of additional programmer time optimizing the software. You can whine and kvetch about this, but it won't change reality.

Back when I was in college and was taking EE/CompE courses I couldn't believe how crappy the x86 ISA was either. And it is crappy. So what? It's still faster than everything else out there, it's cheaper than the competition, and the world has boatloads of software that runs on it. Do you have any idea how much software is used on a daily basis that hasn't been touched in years? How much do you think it would cost to replace all that software?

Don't worry. One day you'll graduate too and after a couple years in the real world you'll discover that a crappy solution that fits the job is far, far better than a perfect solution that doesn't do anything.

Could Using Metal Result in Cheaper Chips?

[mykepredko]

Right now, most chips have the aluminum (interconnect) to Silicon (gate) interconnect, but if the gate was metal, couldn't you replace the need to put down these two materials with a single metal deposition? I am not intimately familiar with the different steps of chip manufacturing, but even if the metal used for the gate was somewhat unusual (ie nickle), wouldn't it make sense to work towards eliminating one of these steps to reduce the overall manufacturing cost?

myke

Re:Could Using Metal Result in Cheaper Chips?

[joto]

I am not intimately familiar with the different steps of chip manufacturing, but even if the metal used for the gate was somewhat unusual (ie nickle), wouldn't it make sense to work towards eliminating one of these steps to reduce the overall manufacturing cost?

No.

What the article was speaking of was using a metal as insulator. Since this doesn't make any sense at all, what they probably meant is a metal-oxide. You don't want an insulator as interconnect, because it doesn't interconnect.

The Apollo process, not Gemma

[TubeSteak]

The Apollo Method (Skip to the end of the article)
1. Place diamond wafers on pedestal. Depressurize chamber to one-tenth of an atmosphere.
2. Inject hydrogen, natural gas (CH4) into chamber. Heat with microwave beam. At 1,800 degrees Fahrenheit, electrons separate from nuclei, forming plasma.
3. Let it rain. Freed carbon precipitates out of plasma cloud and is deposited on wafer seeds.
4. Let it grow. Wafer seeds gradually become diamond minibricks, building up at half a millimeter a day.
5. Open chamber and remove diamond brick. Slice into wafers for semiconductors or cut and polish to make gems.
6. Profit!!!

DeBeers and Co. are very very unhappy about these two technologies and what they're going to do to diamond prices. Both companies can create perfect diamonds and the second manufacturing process will allow (once its been scaled up) for diamonds to be used in electronics.

But here's the reason the U.S. might just end up behind the technology curve:

"Diamonds represent a seismic change in semiconductors," says Krishnamurthy Soumyanath, Intel's director of communications circuits research. "It takes us about 10 years to evaluate a new material. We have a lot of investment in silicon. We're not about to abandon that."

...frustrated with what he thinks of as myopia in the US computer business. "Europe and Japan have been investing in diamond semiconductor research," he says, citing the Japanese government's announcement in December that it would begin allocating $6 million a year to build a first-generation diamond chip...

Also, some other posters have commented on impurities being a stumbling block for diamond-based electronics, how convienent that "CVD diamond precipitates as nearly 100% pure"

Re:The Apollo process, not Gemma

[*weasel]

intel is talking about this new substance being used within 5 years, as opposed to 10-15.

In the end, diamonds may be great, but if they remain more expensive than silicon or metal-oxide, they aren't necessarily the best choice.

nevertheless, when they do hit (and i personally believe they will) intel will certainly buy into the field.

After all, they're more a development shop than a pure research shop. They almost certainly won't be the first to use diamonds in chips, but you can bet they'll be one of the first to refine diamond-chip production and fabrication.

Moore's Law is NOT a Law

[G4from128k]

The changes are largely necessary because of the unsavory consequences of Moore's Law, the famous dictum that states that the number of transistors on a chip doubles every two years

Moore's Law is only an empircal observation -- a convenient curve that fits through the our current data on time and transistor count. There are no gaurantees that this trend will hold for the future.

The point is that no physical phenomena forces the doubling. At best, one could say that mental and procedural limits prevent doubling faster than Moore's so-called Law. Perhaps this is the more interesting Law -- that doubling can't occur faster than every 18 to 24 months.

Re:Moore's Law is NOT a Law

[JGski]

Moore's Law is a market imperative, which to a business is pretty much the same thing as a law.

Already been done...

[fetus]

The TX in Terminator 3 had metal ones instead of silicon...

Re:Already been done...

[Xeger]

Is this just an off-the-cuff remark, or did T3 actually mention this fact to the viewers sometime during the movie? If so, can you remember how it was phrased?

Obviously, the TX was a refinement of and an improvement on the T-1000 and its "gloopy metal" construction. But aside from mentioning that the T-1000 used a decentralized computing architecture, they never got down to specifying where inside all that gloop the computation was taking place...I guess we were supposed to believe that every molecule of the metal was somehow contributing toward the running of the T-1000.

Come to think of it, acting like a T-1000 doesn't take much computing power. I bet my desktop PC could do it! All it needs to do is say "Get out" and "Tell me where this boy is" in a monotone, and randomly stab people. Should be a binch!

Re:Already been done...

[ViolentGreen]

Fabuoulsy metal ones...

Re:Already been done...

[taradfong]

...and even if every molecule did computing, what about power?

Re:Already been done...

[Xeger]

Good point. I think we'll just have to suspend our disbelief here, since with my limited imagination I can't think a valid answer to either the computation or power problems. Looks like no gloopy metal robots for us anytime in the near future!

In fact, the power needed to run the T-1000's computations pales in comparison with the energy required to move all of that metal around. I can think of two ways to get that much power:

1) "Eat" organic matter you find in the environment and metabolize it to produce energy.

2) Carry an onboard power plant, such as a fusion or antimatter reactor.

In either case, individual computation units could be equipped with super-high-density storage batteries, so they could survive away from the power plant for short periods of time. This suggests, of course, that computation units are much bigger than a single molecule.

Re:Already been done...

[taradfong]

Although it wouldn't work for T2, I suppose you could have a central brain and power pod which would get swished around to keep it wishin the goopy part. But I wouldn't invest in a startup that claimed they could do it.

Re:So what they're saying is...

[morcheeba]

No, coppermine wasn't about using metal on gates. It was an all-aluminum (go figure!) interconnect scheme that used a low-k dielectric and thick wires for faster speeds. Only later did copper get used (and then again, only for the interconnects, not the gates)

Re:Alien Technology?

[JUSTONEMORELATTE]

Now how can you say that CPUs are based off of alien technology when Intel is making changes like this?

The same way we've always been saying it -- emphatically

--

Multi processing versus single fast CPU

[goombah99]

Via and transmeta and to a much lesser extent centrino and the G4 have all showed that a better processor LOGICAL architecture can achieve a better power/gflop ratio than conventional x86 chips. Of course they aren't pushing the speed limits for a single processor like the fast intels and poser pc chips do.

but the point is that ratio is actually quite large. via's 7 watt chip is 2 to 3 times slower than intels 300 watt chips. which is about an order of magnitude. Moreover the transmeta designs have shown an increasing ratio of gigflops/watt as their designs evolved.

therefore multi processor designs with say 10 to 100 chips, would use less power than the power4 or intel chips. even if they were only 50% efficient in throughput this would still be vastly faster computer.

the figure 50% might be reasonable. currently most small cpu-count multi-processors do much better than this on code that is suited to multi-processing. Most code as currently written is not. thus we need clever compilers and maybe more expressive computer languages (e.g. steal the brilliant ideas in fortran 2000) which allow the programmer to give hints on how a singly threaded procedure can be tranparently parallelized. THis wont be highly efficient but heck we've got chips to waste: 100 VIAs would only use 700 watts of power.

moreover IBM has long been pointing out that the problem ultimately is not processor speed, its memory access speed. If you think about this for a while you suddenly realize that the ideal computer is one that is 1) accessing a given bank of memory at the memorys maximum possible rate continuously without pausing 2) that every byte that is fetched is always used and never just along for the ride as part of a page load or part of a predictivie lookup that might be tossed out. 3) that the memory should be divided into banks all of which can be fetching simultanouesly. This computer would have many many very slow CPUS each fetching exactly what they need at the moment and nothing else; each bank of memory would be shared by enough slow cpus to saturate the memory request rate--often the CPUS would be waiting on memory but that's okay sinc enothing is wasted.

Re:This might help save the world (for a little wh

[ultranova]

AFAIK the last Ice Age (the North American part, don't know about Europe) was caused by the Gulf Stream. It brought warm water to North Atlantic, which caused plenty of evaporation, and when the resulting warm, moist air spread over the North America and hit the cold air there, the water recondensed and rained down as snow. Snow reflects sunlight more than soil, so this caused the continent to get even colder, which caused more snowfall, and so worth.

Besides, glaciers aren't excatly speedsters... Even if the next Ice Age starts right now, we'll all be long dead before they spread much. And even if worst comes to worst, we could always escape to south of equator. There's no good starting points for glaciers there, so the couldness would just be uncomfortable, not lethal.

Or we could just keep on burning even more coal and oil, to pump up the carbon dioxide levels to help couteract the freezing. Maybe drop a few nukes to the oceans to add water vapor to the mix - but of course, it would also increase rainfall and thus speed up the glaciers (more rain -> more snow), as would global warming in any case...

Re:Copper? Vias AND Traces are copper

[dunc78]

Copper interconnects refer to both the vias and the traces connecting the transistors, not just the vias. The "plugs" that connect one layer to another are called vias. For me check out http://zdnet.com.com/2100-11-518399.html?legacy=zd nn . A quote says "Put simply, copper is a more efficient conductor than aluminum, making it useful in processors because it allows for smaller, thinner interconnects. Interconnects are tiny pieces of wire that connect transistors inside a processor."

Type of Metal Used

[Prince Vegeta SSJ4]

No, it will be great. Now your girlfriend can change size, shape and feel of her implants at will using This Metal

She can even form a knife to cut pizza with.

Re:Type of Metal Used

[Tackhead]

> No, it will be great. Now your girlfriend can change size, shape and feel of her implants at will using This Metal
>She can even form a knife to cut pizza with.

Man who cannot cut mustard, can always lick jar.

(Though a chick with boobies that can cut pizzas would be pretty cool.)

"Chip engine" inefficient for different reasons

[poszi]

the Power consumed by the chip is probably MUCH greater than the power delivered by the chip as heat

No. Except tiny amount of energy emitted via radio waves ALL electric energy in a chip is eventually transformed into heat First Law of Thermodynamics.

However, you are right than only a fraction of the heat can be transformed into work via steam engine. A reversible heat engine that has hot reservoir at 370K and cold reservoir at 300K has maximum efficiency of 1-300/370= 18.9% Efficiency of an engine. Silicon chips are too cold to be an effective heat engine.

Do I smell profit to be made?

[freeweed]

I'm curious, if artificial diamonds are now available to the public, cheap, and indistibguishable from the real thing...

What's to stop me from buying a bunch, setting them, and selling them to jewellers who deal in used and reaping one hell of a profit?

Re:Do I smell profit to be made?

[jaoswald]

By "indistinguishable" they generally mean to someone looking at a ring on a finger, or even under a light microscope. But there are ways to distinguish them in the laboratory, through fluorescence, and so forth. The impurities in natural diamonds are different from the ones in synthetics, so deBeers and dealers will still be able to laser-mark their diamonds as natural, and be able to distinguish forgeries, at least for the time being.

The tough part is convincing brides-to-be to accept only the deBeers mark as evidence of lasting commitment, rather than a synthetic diamond that is potentially bigger, just as pretty, and much easier on the groom-to-be's wallet.

But hey, the point is to make the groom cough up as much as he can afford, or more, right? So deBeers might not have so much threat after all.

Re:Do I smell profit to be made?

[freeweed]

But hey, the point is to make the groom cough up as much as he can afford, or more, right?

Exactly, which is why I added "doesn't expect a ring" to my dating criteria.

Limits my choices severely, but it's amazing how much less shallow people are when they don't equate money == love. I'm certainly happier, and the extra dough is fun for a nice trip for 2 somewhere private :)

No help for pacemakers

[Teahouse]

EM pulses do not affect humans, but I don't know if they would penetrate to a pacemaker. Water (i.e. 90% of a human) is a good general radiation shield, but I don't know how it would shield EMP. My guess is that it wouldn't be enough of a shield to protect it. If a microwave oven can knock out a PM, an EM pulse should do the same.

Most pacemaker patients don't need the PM 24/7. Even if it was affected, most would have time to get treated. Some would likely die, but not everyone with a pacemaker would drop dead.

GaAs is the next big thing!

[JGski]

And has been since the late 1970s!!

One day replacement?

[powerlord]

I agree 100% with your comment, but I can't help but wonder if Linux might actually be a path for freedom from x86. If enough people migrate to it as a platform, then it would potentially open up the ability for Software to be ported to other ISA's to have an increase in market share, assuming that "linux" is supported, and that it software is available for multiple platforms (either as a user "recompilation" or with vendor support).

This certainly isn't a scenario I would picture with MS behind the wheel, and in some ways it would parallel what we've started to see with Unix vendors and shops migrating to Linux (which in turn is supporting multiple ISAs).

I'm not saying this is a panacea, just a possibility :)

Can you say "Intel 432" debacle?

[JGski]

Back in the early 80s Intel had VLIW processor called the 432. It almost destroy Intel's bottom line, and they had the resources to make it work.

(I worked at Intel at the time and still have a complete manual set for the 432!).

Re:say what?

[barawn]

One thing that perplexes me is why VLIW is not more popular. Think about it. Spend a ton of energy *once* at compile time to schedule/optimize the code then just run the parallel ops. Why should the CPU do the work of a compiler at runtime?

Because the CPU has more information than the compiler does - i.e., the Halting problem.

The best example that one could give would be interrupt handlers: how could a compiler ever know when an Ethernet interrupt is going to happen? It can't. Nor can it know the code execution path of anything that checks hardware registers, etc.

Yes, the CPU can't know when an interrupt handler will occur, either. But to it, all it's getting is code. It just tries to schedule it as best it can, and so it *is* going to parallelize some things that a compiler can't.

Look at the space consumed by the VLIW core of the Itanium versus the space consumed by the P4's x86 core, minus the common parts (like the execution engines). Yah, the x86 core is bigger, but it's such a trivial amount of transistors compared to the rest of the chip that it's not that big a deal.

It should also be noted that noting high cycle multiplies and then complaining about core efficiency are two totally different things - the high cycle count for certain instructions is a RISC-ism (it's a 'CISC' instruction, that needs to get broken down), which deals with the instruction set itself (or, in this case, the microarchitectural instruction set), and efficiency is complaining about the combination of code+scheduling logic+parallellization hardware.

The P4 is actually heavily optimized for efficiency, since it's such a high clock speed, and crap like pipeline bubbles really hurt you - hence HyperThreading, for example.

I'd really like to know what the net efficiency (that is, how full does the processor's pipeline stay) of the P4 versus the VLIW designs. I don't think I've seen that anywhere.

The problem with metal

[geekee]

The reason people use polysilicon for the gate now is so they can create the gate before they they do source and drain implants. This way the device is self-aligned, since the gate is part of the source-drain mask, producing a device with right amount of gate and source/drain overlap. When they created the source/drain regions using diffusion, it required high temperature, which melts metal. Now, with ion implanting, they probably can avoid high temperature steps after the metal gate creation to avoid melting the metal.

Check your facts!

[Kommet]

The code name Coppermine had NO relationship with the metal used inside the chip. It was still an Al-on-Si chip, just like Katmai. Tualatin (last P-III core) and Northwood (second P4 core) were the first x86 Cu-on-Si chips from Intel (targeting Mobile/Server and Mainstream markets, respectively).

Additionally, AMD was making Cu-on-Si chips back at the Thunderbird (first "L2 cache on core" Athlon) debut. All cores that came from Fab 30 in Dresden were Cu-on-Si while all cores from Fab 25 in Autin were Al-on-Si. Palomino (first Athlon XP core) was made entirely at Fab 30 and thus all Palomino cores were Cu-on-Si.

IBM has been producing Cu-on-Si cores since 1998 (PowerPC 740, IIRC) and producing Cu-on-SOI cores since 1999 (PowerPC 750). Where do you think AMD got their SOI technology?

The Amiga did it right.

[richie2000]

Even today you see several instruction-set architectures implementing single-instruction stream multiple-data stream (SIMD) integer and floating-point operations for multimedia applications, but we need to take this a few steps further--using a general-purpose processor with several special-purpose hardware units that are optimized and integrated for specific tasks, as shown in figure 11. [...] These special-purpose hardware units will provide orders of magnitude of equivalent general-purpose performance.

Ah, they've discovered the Amiga.

Re:The Amiga did it right.

[taradfong]

Ah, they've discoverd the Atari 800. (yes, same designer). And the designer of the Atari 800's SIO serial bus is on the patent for USB. The old becomes new.

Re:The Amiga did it right.

[voodoo1man]

Ah, they've discovered the Amiga.

And pretty much every main-, mini- and real micro- computer design. I guess the computer industry is finally catching up to the 70s fashion mini-revival of a few years ago.

Re:The Amiga did it right.

[joto]

Ah, they've discovered the Amiga.

And pretty much every main-, mini- and real micro- computer design. I guess the computer industry is finally catching up to the 70s fashion mini-revival of a few years ago.

Yeah, and then a decade or two from now, they will soon find out that producing all these specialized units is a waste, when one general purpose processor can easily do the work of all of them.

Then, a decade or two after that, they will move more and more tasks into specialized circuitry, freeing the CPU to do real computation.

Untill, a few decades later, they realize this is a waste, and start eliminating circuitry, having a powerful general-purpose processor do all the work they previously did.

But then people will realize that by adding specialized processor for certain tasks, they free up computing power from the CPU, giving them potentially higher performance.

And so on...

Don't you realize that there is nothing right or wrong about either approach. It ends up as what is right or wrong for the current generation of technology, and this can soon enough change. The best thing we can do about it is producing modular software that can easily be changed to incorporate the advances in hardware.

Re:The Amiga did it right.

[voodoo1man]

Don't you realize that there is nothing right or wrong about either approach. It ends up as what is right or wrong for the current generation of technology, and this can soon enough change. The best thing we can do about it is producing modular software that can easily be changed to incorporate the advances in hardware.

Don't take me wrong - by equating the computer industry with the fashion industry, I'm not disparaging today's passe and unimaginative hardware designs, rather, I'm disparaging the computer industry itself. But you should really quantify that by "technology", you mean the price/performance ratio of hardware. The accelerated-subsystems approach never really disappeared when PCs started replacing "real" computers - PC companies just started selling SCSI controllers and fancy graphics accelerators. Regarding software, I agree with you completely. But it's very interesting to note that "modular" software won't come from any technological gain (or we'd all switch to programming languages from the 1970s - there's really no argument about the superiority of this approach), but rather from a sociological one - thank RMS for free software.

Re:The Amiga did it right.

[joto]

Then we agree on hardware. But regarding "modular" software, I disagree. I think it comes from both technological gains and sociological ones. Modular OS-kernels, languages suitable for writing generic algorithms, containers, and components, component frameworks, as well as cooperation and dictatorship creates modular software. Which is most important matters little, if either the technological or sociological side is missing, it will not happen.

By the way: grouping "70's languages" together makes no sense at all. "70's languages" include as diverse things as C, Fortran, SmallTalk, COBOL, ML, Icon, CLU, Forth, PL/1, and Prolog. Some of these languages are still alive today. Some of them should be. Some of them shouldn't be. Some of them have worse modern equivalents. Some of them have better modern equivalents. And there is certainly an argument about the "superiority" of going back to using only 70's languages. Amongst other things, you will miss: C++, Java, Perl, Python, Tcl, PHP, C#, Common Lisp, Haskell, and several other languages that I certainly prefer to exist rather than being ignored because they are newer than Jan 1, 1980. In particular, all the "scripting" languages certainly have helped getting closer to the goal of "mularity".

Re:This might help save the world (for a little wh

[PitaBred]

http://www.usatoday.com/usatonline/20031029/563101 1s.htm Or you can reject FUD in all walks of life, not just computers.

Health concerns

[starm_]

With CPU's operating at or above microwave oven frequencies, maybe Intel is just trying to sheild us from the harmfull radiations.
Hey and since the CPU clock is a square have couldn't it even have highger more dangerous harmonics?

Re:say what?

[michael_cain]

Why is VLIW not more popular? Because compiler technology isn't yet good enough...

VLIW seems to have worked out reasonably well in specialized niches -- TI's DSP chips and media processors by Equator Technologies are examples. I know that Equator has been working on their proprietary compiler technology for on the order of 15 years, so your comment about compiler technology is pretty much on target -- the people who seem to have at least some of the answers are holding them rather tightly. And of course, many of those niche applications can afford to spend time hand-tuning critical code sections.

Moore's Law or self-fulfilling prophecy?

[G4from128k]

Moore's Law is a market imperative, which to a business is pretty much the same thing as a law.

Interesting insight. I wonder if there is an accidental collusion among semiconductor companies to limit their progress to Moore's observed trend? It seems suspicious to me that the trend should continue for so long without an obvious physical cause. In my orginal post, I suggested that mental and procedural limits kept companies for doubling faster than Moore's Law -- people just don't seem to create magic breakthroughs that double the transistor count in 3 months.

But now I wonder if Moore's law is a self-fulfilling prophecy. Everyone (semiconductor makers, software creators, and chip customers) knows about the Law, so everyone obeys it. Rather than spend time doubling the transistor count in a very short time, companies stick to the industry trend and spend time on other advances (e.g., innovations in microcode, cache, bus, branch-prediction, etc.)

The point is that in business, you need only beat your competitors by some incremental value. Thus, there is little incentive for Intel, for example, to double transistor count in 6 months as few customers would pay much more for the new breakthrough-density processor than they would for a competition-beating processor that only doubles on an 18-24 month schedule.

Perhaps Moores Law holds because everyone obeys it -- makers are too afraid to go slower and there's little competitive advantage to going much faster.

Re:Moore's Law or self-fulfilling prophecy?

[Mark of THE CITY]

Lasers of shorter wavelength, and steppers of finer intervals, pace transistor size, I believe. Maybe the business types figure that buying the latest-and-greatest (and costly) fab equipment can only be justified at intervals that fit the Moore's law curve.

Re:Moore's Law or self-fulfilling prophecy?

[ahdeoz]

When Intel had a virtual monopoly, they enforced the "law" fairly well. But in the late 1990s, competition from AMD forced them to exceed their delivery targets. Transistor fab advancements do come suddenly, and typicall allow for triple or greater increase in speed, but because it is more lucrative to graduate the curve, and also because it increases stability, the law has held fairly well over time. For instance, the past 3 years have seen markedly slower growth than Moore's law predicts, but averaged with the 3 previous, it still holds.

Re:Moore's Law or self-fulfilling prophecy?

[G4from128k]

For instance, the past 3 years have seen markedly slower growth than Moore's law predicts

Interesting. Do you know why growth has slowed? Is it because of process barriers -- they can't make the feature size smaller? Or is it because of complexity barriers delaying design roll-outs -- designing chips with tens of millions of transistors is too hard? (I would doubt that competition has slackened much or that the effects of the dot-com crash would have hit Intel transistor count roadmap 3 years ago).

You know you're reading too much Slashdot...

[Haeleth]

...when someone posts a chemical formula and you try to read it as leet-speak.

Remember Micro Alloy Transistors?

[HermanZA]

This is very mature technology...

new semiconductor material

[aimew]

I don't know about the metal (as an insulator?) but I'd be on the watch for diamond chips in the not-too-distant future. See:

http://www.apollodiamond.com/about.html

and to a lesser extent:

http://gemesis.com/about.htm

Very high thermal characteristics, radiation hard, much less fagile, optically clear from infra-red through ultra-violet, are among the things that silicon is limited by that are diamond's strengths.

Now they are becoming commercially available as wafers. (Soon anyway.)

Moores(sp?) law is in no danger.

Uh

[autopr0n]

Regardless of your opinion on the above (some of which is highly speculative), this leads us to the vision of a computer technology where not not only electronics states are used for data processing, but magnetic ones as well.

Magnetic feilds come from electrons.

metal Gates?!

[hellmarch]

I knew Bill Gates would make some sort of mechanial exoskeleton and destroy us all someday...

Mularity

[joto]

"mularity" of course being a typo for "modularity"

Re:This might help save the world (for a little wh

[basingwerk]

Oh, that's all right then.

Re:This might help save the world (for a little wh

[basingwerk]

Thanks for your reply. This report (by Nick Shultz of techcentralstation.com) directs us to read a new study in Energy & Environment, which he believes refutes certain data related to global warming. Whatever the facts of this case, I have adjusted my FUD detectors to 'extra sensitive' and I use large pinches of salt with respect to both sides of this discussion. On the other hand, though, when uncertainty exists about important things (like the end of the world), I prefer to play it safe and err on the side of caution - but that's just me.

Re:Revolution Ending Revealed!

[dytin]

How did you come up with all of this??? I applaud you , because I liked your script MUCH more than the movie that I saw. Too bad you didn't work with the Wachowski brothers, if they would have incorporated half of your ideas, the movie would have been ten times better.

Your movie was better because, it explained why neo had psychic powers while still in the real world, it gave a purpose and a reason as to why he was stuck in the train station at the beginning, it had more and cooler fight scenes, Morpheus had a legitimate part, and finally, the ending of your movie is actually a happy ending.

Overall, I think that you should do a re-make and use this script instead.