Intel announced a major breakthrough in microprocessor design Friday that will allow it to keep on the curve of Moore's Law a while longer. IBM, working with AMD, rushed out a press release announcing essentially equivalent advances. Both companies said they will be using alloys of hafnium as insulating layers, replacing the silicon dioxide that has been used for more than 40 years. The New York Times story (and coverage from the AP and others) features he-said, she-said commentary from dueling analysts. If there is a consensus, it's that Intel is 6 or more months ahead for the next generation. IBM vigorously disputes this, saying that they and AMD are simply working in a different part of the processor market — concentrating on the high-end server space, as opposed to the portable, low-power end.
As a graduate student researching this field, this is an amazing bit of news! - The intel high-k announcement is a *major* breakthrough, and a new, disruptive technology for chip technology especially as far as the the introduction of new materials in the Fab are concerned (and trust me, Fab engineers are paranoid about such kinds of shifts). It essentially involves replacing the SiO2 dielectric gate insulator with a new class of materials, very likely Nitrided Hafnium Silicates (though they have not publicly acknowledged the silicate part, they just mention it as a compound of Hafnium - it is the leading contender in the field).
The high-k film can be made physically thicker than the very thin SiO2 layer (which is only around 12 Angstroms thin at the moment, making it leak like a sieve) without messing up the capacitance requirements for the transistor. The introduction of new metal gate instead of the classic poly-crystalline silicon (called poly) is also abig deal, and there is greater secrecy on what those materials are. The wikipedia article on high-k has the details. http://en.wikipedia.org/wiki/High-k_Dielectric [wikipedia.org]
Yes, Moores' law didn't account for dupe postings. If we could just post this news a few more times today we could jump decades ahead in terms of transistor density! Keep up the pace dear editors:)
Sorry but why is this being reported again now? We already knew Intel and IBM had achieved a 45nm process and that it would be coming to mass-market chips in 2007-08. It's 2007 and it's here. Hooray and all that, but is a company following through on its claims really so shocking that it constitutes being reported again... twice [slashdot.org]?
pretty sure this article was more about the switch to Hafnium as an insulator as apposed to the 45nm technology. Also the fact that they are using a new silicon substrate over the existing standard...
The hafnium and high-k metal gates are pre-requisites for the 45nm process. The two articles highlighted might vary somewhat in focus but they're definitely reporting the same thing.
Well whether they use that particular method or not, the point is that the existing materials Intel are using for the 65nm process apparently aren't up to the task at the 45nm scale. If that's wrong well I guess I've been misled by the articles I've read on the subject.
The alternative would have been just to shrink the devices, gain less on performance and use circuit techniques to battle parasitic power consumption. That is what most companies in cost sensitive markets are going to do.
But how much further will that get them before RFI makes it a moot point? At that small of a pathway, I'd think that random radio signals and electrical noise would be disastrous.
Also, how well does this survive long term? Is it resistant to electromigration [wikipedia.org] over time?
All great to hear, but I'm not sure how long this will let them keep pace with Moore's law, at best it buys a couple more years of progress on current processor designs I guess.
Good point. That is the very reason NASA sticks to 386 and earlier vintage computers from what I have read. Outside of the insulating atmosphere, cosmic rays pass throuh and have a tendency to be larger then the cicuit gap. This makes for some interesting and adverse additions to any computation. Every now and then the normal press reports new advances in biological comuters, light based, heck I even read of a wooden one once... Nothing it seems ever comes of it though exce
NASA sends up several thinkpads with every shuttle launch. They have been doing so since 1993, and they have upgraded many times to more modern machines. STS-114 was the first to fly several A31p thinkpads with 1.8Ghz p4s.
For the mars missions and things like it, radiation hardened processors like the RAD750 are used. It seems that everything in use is at least pentium class.
No the shuttle and station run on older stuff because those processors are radiation immune, and they are critical systems that cannot crash. The laptops are for everyday work that do not interface to the shuttles systems. If they crash from the radiation, the astronauts simply put it aside and grab another one.
The shuttle used hand woven magnetic core memory until 1990.
Yep. Stable, information-retaining (unfortunately, it even retains info after immersion in seawater), and basically immune to cosmic ray disruptions. Which doesn't require a lot of error-correction circuitry.... Not terribly data-dense or fast compared to semiconductor (part of the reason to replace it, after all) but it works.
It was designed in the 60s...
Actually, the computers themselves were designed the 70s, with updates in the 80s; core mem
The shuttle internal systems run on obselete crap.
Obselete, incredibly reliable, utterly adequate rock-solid gold. If it ain't broke, don't fix it. Launching enormous rockets with software control is possible to screw up [wikipedia.org]. Given the choice, I'd rather fly with the proven computers.
There are many, many people spending their careers solving those types of problems.
It's not really interesting when someone does something in 45nm. It's interesting when enough of the problems with 45nm are solved for it to actually be practical to make 45nm-based chips.
So, the answer to your question is: someone figured it out already.
Electromigration is only an issue at high current densities. For clarification, "high" is defined as the density where electromigration becomes an issue. The solution is use less current, use more metal so the current is less dense, or find a material that can handle higher current density.
It depends on whether the engineers do their jobs and check their current density. If they do, no problem. If not, some percentage of the chips with eventually fail. Worrying about it doesn't help. Engineers checking it is the only thing that helps.
Testing for electromigration issues is standard operating procedure for companies like Intel. They basically pump insanely high amounts of current densities through their devices and see how long they take to fail. Then, they can use that figure to extrapolate how long they'll take to fail under normal conditions. Basically, they can test years worth of damage in days. Asking whether Intel checks for this is like asking whether car companies check to see if the engines start up before selling them. Of cours
Welcome to the club! On your application as editor, did you have to swear that you don't actually read slashdot as a precondition for employment like all the other editors?
Hey now, you should be positively thanking him. The previous posting had an awful summary that didn't mention IBM, AMD, or the fact that the new High-K replacement was based on hafnium (they misspelled it as halfnium in the actual article, which was even worse).
At least with this summary we'll get cool arguments about Intel vs. AMD and IBM and conspiracy theories and stuff.
That's no misspelling, it is halfnium! You could have understood this yourself, if you hadn't been so quick to dole out criticism, and instead had spent a second considering the fact that they reduced the size from 90 nm to 45 nm.
apart from the dupe, kdawson is possibly the best editor they have. I for one, blame our new firehose overlords - so its our fault for voting for 2 of the many posts about this news.
The Intel announcement is new evidence that the chip maker is maintaining the pace of Moore's Law, the technology axiom
I thought it's an empiric law; the definition of axiom is quite different from that.
Intel said it had already manufactured prototype microprocessor chips in the new 45-nanometer process that run on three major operating systems: Windows, Mac OS X and Linux.
Again, I thought it's the operating systems who run on microprocessors, not vice-versa. And I [not being a kernel developer, though] can't see any reason for an OS to stop functioning on a new processor model if the architecture is intact and no serious hardware-level bugs are introduced.
Intel said it had already manufactured prototype microprocessor chips in the new 45-nanometer process that run on three major operating systems: Windows, Mac OS X and Linux.
Again, I thought it's the operating systems who run on microprocessors, not vice-versa. And I [not being a kernel developer, though] can't see any reason for an OS to stop functioning on a new processor model if the architecture is intact and no serious hardware-level bugs are introduced.
The OS'es running on the prototypes is probably meant to show that there are functioning processors made using the new process, as opposed to a couple transistors in a lab. A kind of proof this isn't just vaporware to boost stock prices.
Probably not. From my understanding of this new tech, silicon will remain the substrate of the chips. Hafnium is only used as an insulating material layered on the top. So the quantities of hafnium will be extremely small in relation to the amount of silicon. Along with a smaller (45nm) process, the total amount of hafnium in a single chip should be quite small.
If there is a consensus, it's that Intel is 6 or more months ahead for the next generation. IBM vigorously disputes this, saying that they and AMD are simply working in a different part of the processor market
Didn't read TFA, but is it possible to have a consensus with one party vigorously disputing it?
It would seem to be consensus of the analysts, but who knows how accurate that is if one company is disputing the information leading to the consensus.
The funny thing about this is that every few weeks you read some article that says, "Yup! That's it! We simply cannot get any more out of Moore's Law! It's dead."
Then a couple weeks later someone says, "Yup! We're gonna squeeze a few more years out of Moore's law. New advance! It isn't dead!"
Moore's Law is like the Energizer Bunny. It just keep's going.
If you change the definition to something Moore never said, then of course it might not fit. Moore's original definition had to do with number of transistors on "an integrated circuit." The original graph didn't even specify size. (Goto Intel.com, search Moore's Law, all that stuff is there.) That engineers have been unable to keep up with exploiting the law isn't really all that surprising or uncommon. However, the interesting thing about Moore's Law is that if you extrapolate the graph backwards in time, i
Well, it's about time. Hafnium oxide dielectrics were the talk of the semiconductor research world in the early/mid 90's. Big-time chip manufacturers refused to adopt the technology, though, hoping that some technology that didn't require the re-vamping of an entire fabrication facility would come along and magically reduce gate oxide lekage current.
The technology is fairly mature by now (from a research standpoint), so the only "news" is that the major manufacturers have finally realized that it is th
This article's summary is far more accurate and informative than the other one. I posted several times in the older post to help clear up some misinformation (the article it linked to misspelled hafnium as "halfnium" and only mentioned it once, and never mentioned IBM or AMD).
Actually I believe there are only two properties of diamond that are superior to silicon in respect to electronic application: Heat conductivity and band gap.
The disadvantages are numerous, starting with the very basic fact that there is no known n-type dopant for diamond.
Two breakthroughs in one day? (Score:5, Insightful)
This is a big deal (Score:5, Interesting)
The high-k film can be made physically thicker than the very thin SiO2 layer (which is only around 12 Angstroms thin at the moment, making it leak like a sieve) without messing up the capacitance requirements for the transistor. The introduction of new metal gate instead of the classic poly-crystalline silicon (called poly) is also abig deal, and there is greater secrecy on what those materials are. The wikipedia article on high-k has the details. http://en.wikipedia.org/wiki/High-k_Dielectric [wikipedia.org]
Parent
Re:Two breakthroughs in one day? (Score:5, Funny)
Parent
Re: (Score:2)
Moore's "Law" isn't - it's more a rule-of-thumb.
Not news (Score:4, Insightful)
Re: (Score:2, Insightful)
Re: (Score:2)
Re: (Score:3, Insightful)
That is not true. There will be a number of companies doing 45nm without high-k and metal gates.
Re: (Score:2)
Re: (Score:3, Informative)
The alternative would have been just to shrink the devices, gain less on performance and use circuit techniques to battle parasitic power consumption. That is what most companies in cost sensitive markets are going to do.
Chip Breakthrough.... (Score:3, Funny)
RFI? Electromigration? (Score:2, Insightful)
Also, how well does this survive long term? Is it resistant to electromigration [wikipedia.org] over time?
All great to hear, but I'm not sure how long this will let them keep pace with Moore's law, at best it buys a couple more years of progress on current processor designs I guess.
Re: (Score:2)
Every now and then the normal press reports new advances in biological comuters, light based, heck I even read of a wooden one once... Nothing it seems ever comes of it though exce
Re: (Score:2)
For the mars missions and things like it, radiation hardened processors like the RAD750 are used. It seems that everything in use is at least pentium class.
Re:RFI? Electromigration? (Score:4, Informative)
Parent
Re: (Score:3, Insightful)
Yep. Stable, information-retaining (unfortunately, it even retains info after immersion in seawater), and basically immune to cosmic ray disruptions. Which doesn't require a lot of error-correction circuitry.... Not terribly data-dense or fast compared to semiconductor (part of the reason to replace it, after all) but it works.
It was designed in the 60s...
Actually, the computers themselves were designed the 70s, with updates in the 80s; core mem
Re:RFI? Electromigration? (Score:4, Insightful)
Parent
Re:RFI? Electromigration? (Score:4, Interesting)
It's not really interesting when someone does something in 45nm. It's interesting when enough of the problems with 45nm are solved for it to actually be practical to make 45nm-based chips.
So, the answer to your question is: someone figured it out already.
Electromigration is only an issue at high current densities. For clarification, "high" is defined as the density where electromigration becomes an issue. The solution is use less current, use more metal so the current is less dense, or find a material that can handle higher current density.
Parent
Re: (Score:2)
Re:RFI? Electromigration? (Score:4, Funny)
Parent
Re: (Score:2)
Is this kdawson's first front page dupe (Score:4, Funny)
Re: (Score:2)
At least with this summary we'll get cool arguments about Intel vs. AMD and IBM and conspiracy theories and stuff.
Re: (Score:3, Funny)
Re: (Score:2)
printer/ad free version (Score:3, Informative)
Axiom? (Score:5, Insightful)
I thought it's an empiric law; the definition of axiom is quite different from that.
Again, I thought it's the operating systems who run on microprocessors, not vice-versa. And I [not being a kernel developer, though] can't see any reason for an OS to stop functioning on a new processor model if the architecture is intact and no serious hardware-level bugs are introduced.
Re: (Score:2)
Well, yeah. That's pretty much the point. Usua
Re: (Score:2)
Rename? (Score:5, Funny)
Re: (Score:2)
I know silicon is a pretty common element, how difficult is it to find hafnium? If it is rare, could this lead to super expensive chips?
Re: (Score:2)
Re: (Score:3, Funny)
Let's hope that real estate prices get cut in haf :(
--Rob
Whaa? (Score:4, Insightful)
Didn't read TFA, but is it possible to have a consensus with one party vigorously disputing it?
'course! (Score:2)
Re: (Score:2, Insightful)
Moore's Law is Dead! Or not! (Score:4, Funny)
Then a couple weeks later someone says, "Yup! We're gonna squeeze a few more years out of Moore's law. New advance! It isn't dead!"
Moore's Law is like the Energizer Bunny. It just keep's going.
Re: (Score:3, Funny)
Moore's Law is like the inappropriate apostrophe. It just won't die.
Re: (Score:2)
Moore's original definition had to do with number of transistors on "an integrated circuit." The original graph didn't even specify size. (Goto Intel.com, search Moore's Law, all that stuff is there.) That engineers have been unable to keep up with exploiting the law isn't really all that surprising or uncommon. However, the interesting thing about Moore's Law is that if you extrapolate the graph backwards in time, i
If you're into investing ... (Score:2)
It's Da Bomb! (Score:2)
Finally... (Score:2, Interesting)
The technology is fairly mature by now (from a research standpoint), so the only "news" is that the major manufacturers have finally realized that it is th
Re: (Score:3, Informative)
Re:How long for this to reach laptops? (Score:4, Insightful)
Parent
Re: (Score:2)
2) Next year
3) More expensive
4) How much are you willing to spend?
Re: (Score:2)
Re: (Score:2)
Actually I believe there are only two properties of diamond that are superior to silicon in respect to electronic application: Heat conductivity and band gap.
The disadvantages are numerous, starting with the very basic fact that there is no known n-type dopant for diamond.
Re: (Score:3, Interesting)
Phosphorus http://www.aist.go.jp/aist_e/latest_research/2005/ 20050615/20050615.html [aist.go.jp]
Re: (Score:2)
Nice, I was not aware of the later work. It is still a far way towards proper junctions.
Re: (Score:2)