Sign in with
Sign up | Sign in
Your question

Update: Intel testing 32nm, 22nm 2015?

Tags:
  • CPUs
  • IBM
  • Intel
  • Product
Last response: in CPUs
Share
December 6, 2006 1:32:49 PM

Just read this, 32nm on track, but 22nm might not be here until 2015. I just wonder if that is where AMD will catch up on the process size front? The reason I ask this is because of the cooperation they have with IBM, and IBM is pretty inovative. I can't find anything on how IBM is doing 22nm fabrication, other than they are working on it. Any thoughts on this people?

wes

http://www.digitimes.com/bits_chips/a20061206PD207.html

More about : update intel testing 32nm 22nm 2015

December 6, 2006 1:46:04 PM

Quote:
Just read this, 32nm on track, but 22nm might not be here until 2015. I just wonder if that is where AMD will catch up on the process size front? The reason I ask this is because of the cooperation they have with IBM, and IBM is pretty inovative. I can't find anything on how IBM is doing 22nm fabrication, other than they are working on it. Any thoughts on this people?

wes

http://www.digitimes.com/bits_chips/a20061206PD207.html


AMD can't compete with Intel on the fabrication technology front due to the capital constrain. Although IBM is innovative, AMD just licenses IBM's technology, not co-develops with it.
December 6, 2006 2:14:06 PM

Well, my point is, if the 45nm node is supposed to be out from Intel in 08 at the latest, and lets just say 32nm in 09-2010..... then 2015 for the following node..... you don't think AMD/IBM could possibly catch up? And according to this they are co-developing.

"AMD, IBM Co-Developing Chipmaking Technologies

Sunnyvale, Calif.-Advanced Micro Devices of Sunnyvale, and IBM will jointly develop technologies for producing future high-performance ICs with improved microprocessor performance and lower power needs.

The processes involved are based on advanced IC structures and materials, including high-speed SOI transistors, copper interconnects and improved low-k dielectric insulation.

The AMD-IBM agreement targets the use of 45nm and 65nm lithography on 300mm wafers. The development is supported by engineers from both companies working together at IBM's Semiconductor Research and Development Center in East Fishkill, N.Y.

The first products based on the new 65nm technologies are expected to appear in 2005, according to the companies. [amd.com] [ibm.com]"

http://www.chipscalereview.com/issues/0303/industry_new...

Also, AMD is doing better today than the were two years ago. They are building new FABs and such. I am just curious what people think.

Thoughts?

wes
December 6, 2006 2:23:38 PM

AMD will never scale faster than Intel just based on past with Intel's speed at shrinking and AMDs troubles they've had. AMD has not gone through die shrinks as successful as Intel. Point in case AMDs transition to 65nm has been a rocky one, rumored and looks to be true since its taken them awhile to get them out. Their transition to ever smaller die sizes will just get harder. And on top of all of that Intel's production capacity allows for it to to the speed that it develops its chips while AMD shall i say is FAB bottlenecked :wink:
December 6, 2006 2:29:11 PM

Historically you are correct. But, the partnership with IBM is still maturing. And, again, I am not talking soley about AMD, I am talking about AMD and IBM working together, coupled with the fact that Intel is saying the die shrink from 32nm to 22nm might take 6-7 years. Along with the fact that AMD IS building more FAB's. See my point?

wes
December 6, 2006 2:32:19 PM

Now I aint no quantum physics egghead, but at 22nm aren't electrons start going through silicon, showing up on the other side, duplicating themselves, doing the mambo and generally introducing Mr. Heisenberg into places where he don't belong? :wink:
December 6, 2006 2:37:00 PM

There are many physical limitations - some that we just don't understand once you get down past 32nm. We think the answer is to step away from silicon and start using wafers made of other substances.

Kinda funny that article you linked says amd will have 65nm chips out by 2005 :) 
December 6, 2006 2:42:28 PM

Quote:
There are many physical limitations - some that we just don't understand once you get down past 32nm.


At less than 10% the width of UV-C there's a whole lot we don't understand. How can any company announce with a straight face that they are going to release something at a point in the future when that something may be physically impossible? I'll just run a press release stating that my company is going to release in 2016 a CPU built on a 5nm process and I want to attract $100 billion in VC cash to build the factory. Doesn't matter that at that scale, electrons are the size of standing wave boulders. I'll bet some Silicon Valley shyster can hook me up with some dumba$$ VC and get me the money...

Christmas is coming, so... bah humbug! :roll:
December 6, 2006 2:46:00 PM

Quote:
Historically you are correct. But, the partnership with IBM is still maturing. And, again, I am not talking soley about AMD, I am talking about AMD and IBM working together, coupled with the fact that Intel is saying the die shrink from 32nm to 22nm might take 6-7 years. Along with the fact that AMD IS building more FAB's. See my point?

wes


The catch up could happen, but seems to me that the change from 32 to 22nm might be harder than the previous die shrinks, also you need to take account that Intel might be bringing new transistor technology (maybe 3d transistors) and optical links into it's processors around the time they go 22 nm. There're too many variables right now to question this: AMD and IBM partnership might not last 10 years, processor market could see a twist (Intel was pressing VIA to stop creating processors...and VIA is about to separate their processor business, which could be acquired by Nvidia which would be interesting to the market), etc.
December 6, 2006 2:47:34 PM

Quote:
There are many physical limitations - some that we just don't understand once you get down past 32nm.


At less than 10% the width of UV-C there's a whole lot we don't understand. How can any company announce with a straight face that they are going to release something at a point in the future when that something may be physically impossible? I'll just run a press release stating that my company is going to release in 2016 a CPU built on a 5nm process and I want to attract $100 billion in VC cash to build the factory. Doesn't matter that at that scale, electrons are the size of standing wave boulders. I'll bet some Silicon Valley shyster can hook me up with some dumba$$ VC and get me the money...

Christmas is coming, so... bah humbug! :roll:

They can make such statements because they have history on their side. A year ago, 45nm was physically impossible. Yet Intel knew that they could do it and have since displayed it working. That is the beauty of the industry, everything that we get tomorrow was physically impossible today.
December 6, 2006 2:49:12 PM

The smaller the die shrinks the longer it will take to develop the process and FABs so i'm not surprised by the time table Intel has put up. AMD hasn't even put up anything that i know of so i doubt they'll get their before Intel.
December 6, 2006 2:56:08 PM

Quote:
That is the beauty of the industry, everything that we get tomorrow was physically impossible today.


Yeah, but there are limits even in quantum physics.

The End Of The Road For Silicon? - Nature Article

Once you butt up against the basic dimensions of atomic structures, there aint nowhere left to go.

I keep telling y'all. Isolinear chips in a Positronic matrix is the way to go. Now if you'll excuse me, I gotta go take my Dilithium... :lol: 
December 6, 2006 2:56:40 PM

Well, it's hard telling what will be going on in 2015. By then, AMD might have come up with a completely new architecture (K20 ;) ) that scales much better. A good example: nobody in 1995 knew that in 2006 Intel would come up with the Core 2. Know what I mean?

So, I think speculation on this is a bit tough.

But, yeah, Intel has done a good job with the shrinkage. I'll give you that.
December 6, 2006 2:59:19 PM

Quote:
But, yeah, Intel has done a good job with the shrinkage. I'll give you that.


Intel's done a better job of shrinkage than jumping in an ice lake after a sauna. But I'm still of the opinion that they are gonna have to go optical or something else since the shrink has shrunk about as much as it can.
December 6, 2006 3:07:13 PM

Quote:
But, yeah, Intel has done a good job with the shrinkage. I'll give you that.


Intel's done a better job of shrinkage than jumping in an ice lake after a sauna. But I'm still of the opinion that they are gonna have to go optical or something else since the shrink has shrunk about as much as it can.
Yeah, it seems like things must move in a new direction. I, personally, am surprised we have coasted so far on silicon. I've been reading those "is it the end for silicon?" articles for a while, but it just keeps ticking along!
December 6, 2006 3:09:09 PM

Problem is going to what? I haven't seen any demos of real, working alternatives that can surpass silicon. So far it's all pie in the sky. Let's just spin some particles here and they'll have the same spin there and we can run Crysis on 'em. Not damn likely. 8)
December 6, 2006 3:10:46 PM

I guess we'll have to start a research company, come up with a new type of CPU, and become billionaires... *sigh*.
December 6, 2006 3:13:50 PM

Quote:
I guess we'll have to start a research company, come up with a new type of CPU, and become billionaires... *sigh*.


You're on. So what if whatever CPU we come up with will be vapourware? Look at the availability of the QuadFX or the 65nm chip. If a big company like AMD can do it so can we. We'll announce a huge leap in computing power, collect the VC billions, have plastic surgery and live out our lives in beachfront mansions in Kiribati.

Sounds good to me. When do we start? :twisted:
December 6, 2006 3:14:37 PM

Quote:
Problem is going to what? I haven't seen any demos of real, working alternatives that can surpass silicon. So far it's all pie in the sky. Let's just spin some particles here and they'll have the same spin there and we can run Crysis on 'em. Not damn likely. 8)


For very high speed chips used for data transmission over 20 GHz they use a Germanium alloy, tought it's really toxic and there're very few manufacturers around.
December 6, 2006 3:18:23 PM

Quote:
For very high speed chips used for data transmission over 20 GHz they use a Germanium alloy, tought it's really toxic and there're very few manufacturers around.


Hey, that's not a bad idea. Get some Germanium Dioxide, use the Pentium D architecture and ship out the chips to all the countries we don't like, North Korea, Cuba, etc. The heat will melt the chips, the toxic fumes will wipe out their livers and kidneys, and we'll have figured out the best way for Intel to get rid of their Pentium D inventory and made the world safe for Democracy! :lol: 
December 6, 2006 3:32:29 PM

Quote:
I guess we'll have to start a research company, come up with a new type of CPU, and become billionaires... *sigh*.


You're on. So what if whatever CPU we come up with will be vapourware? Look at the availability of the QuadFX or the 65nm chip. If a big company like AMD can do it so can we. We'll announce a huge leap in computing power, collect the VC billions, have plastic surgery and live out our lives in beachfront mansions in Kiribati.

Sounds good to me. When do we start? :twisted:
Right now will work...

<Breaking News>
Mr. CaptRobertApril and dean7 have come up with a new kind of processor. It's rumored to be made out of an organic substance that shatters the barriers previously imposed by silicon (think OLED vs LCD). The performance is said to "really kick AMD and Intel in the junk (when it comes out, which is in the future, but right now it can't boot Windows)"

When asked about this technology being vaporware, Dean and Robert replied "hell no, this isn't vaporware! Any idiot can see that it's in our roadmap for 2015!", then they held up what appears to be a picture of their processor next to an Intel quad-core (much like AMD did recently), which clearly proves that this is not vaporware and will smoke Intel.
</Breaking News>

:lol: 
December 6, 2006 3:37:03 PM

Yeah, I know, I saw that one..... thought it was funny also, and they BARELY made 2006.

wes
December 6, 2006 4:38:27 PM

Quote:

<Breaking News>
Mr. CaptRobertApril and dean7 have come up with a new kind of processor. It's rumored to be made out of an organic substance that shatters the barriers previously imposed by silicon (think OLED vs LCD). The performance is said to "really kick AMD and Intel in the junk (when it comes out, which is in the future, but right now it can't boot Windows)"

When asked about this technology being vaporware, Dean and Robert replied "hell no, this isn't vaporware! Any idiot can see that it's in our roadmap for 2015!", then they held up what appears to be a picture of their processor next to an Intel quad-core (much like AMD did recently), which clearly proves that this is not vaporware and will smoke Intel.
</Breaking News>

:lol: 


Now you're cookin'! This is definitely the way to go. To avoid some unpleasant entanglements with the SEC please refer to me as "an experienced, but anonymous inventor". And change "can't boot Windows" to "couldn't run two lines of BASIC." There. I think we have covered our a$$es enough. Ok, so, where's my big VC check? 8)
December 6, 2006 5:21:35 PM

Quote:
Now I aint no quantum physics egghead, but at 22nm aren't electrons start going through silicon, showing up on the other side, duplicating themselves, doing the mambo and generally introducing Mr. Heisenberg into places where he don't belong? :wink:


Not really. An atom is measured in pm. Silicon has an atomic radius of around 100 pm which means that a single silicon atom is about 0.1 nm in diameter. Subatomic particles like protons, neutrons, and electrons diameters are measured in fermi's or fm which is about 1 million times smaller than a nm. So what you would see with 22 nm process is the transitors would be around 200 atoms wide give or take which is still outside the range of Mr. Heisenberg showing his face.
December 6, 2006 5:23:24 PM

Making processors out of tofu has got to be the answer.
December 6, 2006 6:11:53 PM

Quote:
Not really. An atom is measured in pm. Silicon has an atomic radius of around 100 pm which means that a single silicon atom is about 0.1 nm in diameter. Subatomic particles like protons, neutrons, and electrons diameters are measured in fermi's or fm which is about 1 million times smaller than a nm. So what you would see with 22 nm process is the transitors would be around 200 atoms wide give or take which is still outside the range of Mr. Heisenberg showing his face.


200 atoms? Damn! Given the fact that I'm still not fully convinced that anyone on Earth fully understands what an atom is, that is a pretty thin slice of stuff to make your transistors out of. Hell, I still don't get the "electron is a particle and wave" thing, so I think that electron tunnelling is well beyond my grasp and the only thing I understand about Schrödinger equations are that his cat is half-dead. Still, if I have to entrust my life's work that's on my hard drive to something that's not really at any particular point at any particular time, I'm gonna resort to papyrus!

Quote:
Making processors out of tofu has got to be the answer.


Yeah, but then you get into the myriad problems of nigari lithography and then you'll have the Silken Fanboys spreading FUD about the Firm... :twisted:
December 6, 2006 6:24:23 PM

Quote:
Not really. An atom is measured in pm. Silicon has an atomic radius of around 100 pm which means that a single silicon atom is about 0.1 nm in diameter. Subatomic particles like protons, neutrons, and electrons diameters are measured in fermi's or fm which is about 1 million times smaller than a nm. So what you would see with 22 nm process is the transitors would be around 200 atoms wide give or take which is still outside the range of Mr. Heisenberg showing his face.


200 atoms? Damn! Given the fact that I'm still not fully convinced that anyone on Earth fully understands what an atom is, that is a pretty thin slice of stuff to make your transistors out of. Hell, I still don't get the "electron is a particle and wave" thing, so I think that electron tunnelling is well beyond my grasp and the only thing I understand about Schrödinger equations are that his cat is half-dead. Still, if I have to entrust my life's work that's on my hard drive to something that's not really at any particular point at any particular time, I'm gonna resort to papyrus!

The biggest problem they're going to have with going to 22nm is with the lithography. They're getting really close to using X-ray lasers to do the lithography (10 nm to 0.1 nm wavelength). So, while we're not at the limit just yet, we're starting to get close. This is probably the reason they're starting to look at tri-gates instead of the standard binary gate to increase computing power.
December 6, 2006 6:30:43 PM

Trigate. Now that makes a whole lot more sense. I'm only concerned about the fact that it's a concept that's been around for decades but nobody seems to have made it work in a production environment. But at least a trigate doesn't rely on ever-decreasing dimensions down into infinity!
December 6, 2006 7:16:55 PM

Quote:
They're getting really close to using X-ray lasers to do the lithography (10 nm to 0.1 nm wavelength). So, while we're not at the limit just yet, we're starting to get close.

Getting really close? Just to put this in perspective, the wavelength of yellow light is around 570 nm. I don't know what they're using for lithography now, but at a feature size of 65nm it's clear the industry has "long ago" moved past using visible light.

We're already manufacturing objects which can never be "seen" with the human eye. Not merely because they're small, but because they are so small that visible light is too crude and clumsy an instrument to probe them with.

Take a moment to exhale ....

-john, the redundant legacy dinosaur mulling nostalgically about stone ages past
December 6, 2006 7:24:44 PM

Quote:
They're getting really close to using X-ray lasers to do the lithography (10 nm to 0.1 nm wavelength). So, while we're not at the limit just yet, we're starting to get close.

Getting really close? Just to put this in perspective, the wavelength of yellow light is around 570 nm. I don't know what they're using for lithography now, but at a feature size of 65nm it's clear the industry has "long ago" moved past using visible light.

We're already manufacturing objects which can never be "seen" with the human eye. Not merely because they're small, but because they are so small that visible light is too crude and clumsy an instrument to probe them with.

Take a moment to exhale ....

-john, the redundant legacy dinosaur mulling nostalgically about stone ages past

They've been using UV lasers for a while, but as they get closer to X-ray, they require more exotic items to make the lasers work. And unless there is some major breakthrough in particle physics, X-ray lasers will probably be the last lasers used for lithography as the bottom end of the scale for X-ray wavelenght is typically atomic diameter sizes (100 pm).
December 6, 2006 7:55:45 PM

Quote:
http://www.eetimes.com/news/semi/showArticle.jhtml?arti...
EUV making slow progress, says Intel

So how does that tie in, considering that Intel roadmaps have previously said EUV with 32nm?


Historically, Intel has gotten more out of each generation that they had originally planned, so it stands to reason that they could hit 32nm without EUV.
December 6, 2006 8:38:24 PM

IBM and AMD are at the same spot technologically. Together they don't invest as much in research as Intel. I don't expect them to catch up.

I don't expect them to fall further behind either. Being one generation behind Intel has its advantages since Intel does the gruntwork and pushes the silicon industry develop new machines and processes. Everyone else can just buy the machines and can start producing with much less expense. It's not quite THAT easy, but it's not as difficult as being the front runner.

For Intel to be the front runner for so long is a great testament to their engineers and scientists. Those guys know silicon. If Intel is predicting more than 2-3 years for a process shrink generation, then they've spotted some problems that no-one else even knows about.
December 6, 2006 9:16:20 PM

Quote:
Just read this, 32nm on track, but 22nm might not be here until 2015. I just wonder if that is where AMD will catch up on the process size front? The reason I ask this is because of the cooperation they have with IBM, and IBM is pretty inovative. I can't find anything on how IBM is doing 22nm fabrication, other than they are working on it. Any thoughts on this people?

wes

http://www.digitimes.com/bits_chips/a20061206PD207.html


AMD can't compete with Intel on the fabrication technology front due to the capital constrain. Although IBM is innovative, AMD just licenses IBM's technology, not co-develops with it.

what on earth gave you that idea dude, 2015 is a long time god I could catch up to 22nm by that time, i bet you were the sceptic when the AMD atlon kicked the P4's ass!
December 6, 2006 11:12:32 PM

Well if you could catch up to 22nm then i'll be waiting January 1st 2015 to see what you've got to offer since it doesn't even look like we could be using silicon at that stage

your random knock on qcmadness could have been the worst comment in this thread. More and more problems arise as the size gets smaller and smaller so when they get to the very edge it takes longer to develop and will get to a point where the cost to die shrink will outweigh the performance gained, at that point i'm sure intel and amd will have found other materials to use

One thing to consider is diamonds, there has been previous talk of diamond being used theoretically allowing a speed of over 10Ghz and higher. not sure how practical this would be but i know intel has atleast looked into it.
December 6, 2006 11:50:26 PM

Quote:

what on earth gave you that idea dude, 2015 is a long time god I could catch up to 22nm by that time, i bet you were the sceptic when the AMD atlon kicked the P4's ass!



I have serious doubts as to your ability to make toast by 2015. Much less an operational 22nm node logic device.
December 7, 2006 12:16:43 AM

Quote:
Well if you could catch up to 22nm then i'll be waiting January 1st 2015 to see what you've got to offer since it doesn't even look like we could be using silicon at that stage

your random knock on qcmadness could have been the worst comment in this thread. More and more problems arise as the size gets smaller and smaller so when they get to the very edge it takes longer to develop and will get to a point where the cost to die shrink will outweigh the performance gained, at that point i'm sure intel and amd will have found other materials to use

One thing to consider is diamonds, there has been previous talk of diamond being used theoretically allowing a speed of over 10Ghz and higher. not sure how practical this would be but i know intel has atleast looked into it.


Diamond is a very cool material to build electronics out of. One of the best things is that it is the best thermal conductor known to man. About 5 times better than Cooper. It is radiation hardened by the nature of the material. It can work at 1000C with not to much problems. I don't know what its electron or hole mobility is so I can say it can be clocked to reach 10GHz.

Diamond is just a cool material. :D 
December 7, 2006 4:40:47 AM

A couple of questions.

UV lasers? The shortest UV frequency band should be C and that's 100 to 280nm. According to my logic, that should be unable to even work for a 90nm process, a bit like soldering a motherboard with an oxyacetylene torch. What am I missing here?

I do know a bit about diamonds. I've certainly bought enough of them and then watched them walk out the door with the rest of my stuff. And what I do know is that for a crap G colour VVS2 rock not even reaching a carat you're gonna cough up a grand wholesale. Industrial diamonds go up exponentially from this. If you're gonna start computing with these damn things, is Bill Gates gonna be the only guy who can afford it?
December 7, 2006 4:52:32 AM

What about synthetic diamonds?
December 7, 2006 5:19:09 AM

Quote:
What about synthetic diamonds?


I'd think that HPHT synthetics would be out but a CVD maybe boron and deuterium doped would be functional. However, you're still looking at several hundred or even thousands a carat.

I keep telling you all but you don't listen: Dilithium Crystals! 8)
December 7, 2006 5:32:15 AM

At this point in time, I believe that IBM has run transistors at the smallest node. Some time late in 05, they were able to get a functioning transistor on 29 nano tech.
Does this mean that AMD has an advantage on smaller nodes? No.
The main reason why AMD has been slower to the gate (bad pun) on size reduction, has more to do with production than R&D.
AMD has limited production capability, but also limited demand.
As a result, they take thier time with testing. If they are too fast to a node, they will suffer more with lower yields. Then, once the process is mature, they will have an increased yield of 70%. What then would they do with the extra chips?
For AMD, node shrinks are a timing thing.
AMD is now looking at having three fully functional fabs up by around 2010. At that point, they may, or may not be interested in being first to node. I rather believe that they still wont be in any rush. It may seem like a big deal to be behind, but it's probably better to take the time, and save the wafers.
December 7, 2006 8:53:44 AM

Light inversion my man, light inversion.
December 7, 2006 9:08:03 AM

Quote:
Light inversion my man, light inversion.


That's just ancient mid-21st century tech, dude. Nothing beats a positronic matrix with isolinear chips and neural gelpacks. 8)
December 7, 2006 9:41:11 AM

AMD and IBM have learnt much from trying to rush out 65nm parts.

It is unlikely they will make these mistakes on any future die shrinks. (Different mistakes more likely).

Best of luck to both parties, and anyone else who tries.

http://moneycentral.msn.com/stock_quote?Symbol=AMD
http://moneycentral.msn.com/stock_quote?Symbol=INTC
http://moneycentral.msn.com/stock_quote?Symbol=SUNW
http://moneycentral.msn.com/stock_quote?Symbol=NVDA

Sometimes these guys have tech news, or news that most techs can read from a tech point of view, a good week or two before IT websites.

Other times they can be 6 months behind.

It 'pays' to watch the news associated with share prices in this subject IMHO.

8)
December 7, 2006 9:57:52 AM

Diamonds can be made synthetically for industrial use, thus they can cost less with economy of scale.
_________________

8)
December 7, 2006 5:53:49 PM

Quote:
We've got an 11-nanometer transistor working in the lab. That's five generations out, so we can see what's coming. - Pat Maloney

It's near the end of the article: Fortune Magazine article
!