Slashdot Mirror
IBM Leapfrogs Intel With 22nm Chips
Slatterz writes "Intel may be touting 45nm CPUs, but IBM says it can go much further with a strategy to produce future chips using a 22nm fabrication process. The company is adopting a technique called 'computational scaling' in order to manufacture circuits small enough to deliver more powerful and energy-efficient devices. Intel plans to introduce 32nm chips in 2009, but chipmakers have hit a problem in that current lithographic methods are not adequate for designs as small as 22nm owing to fundamental physical limitations. IBM claims to have solved this problem." Unfortunately the phrase "computational scaling" doesn't actually convey any information about how they've solved it.
If I figured out how to do something that would lay a serious hurting on my competition, I wouldn't exactly go around saying how I did it either.
Instead of just saying they're going to do it.
Talk is cheap.
I know its getting harder and harder, especially considering these things are only a handful of atoms across, but why can't they ever skip a generation? Why work on three generations of chips simultaneously? Why not just skip one?
--
find my ip
FTFA: "IBM said that computational scaling overcomes these limitations by using mathematical techniques to modify the shape of the masks and the characteristics of the illuminating source used to image the circuits for each layer of an integrated circuit."
That gives you an idea. They are not being more secretive than normal.
"Civis Europaeus sum!"
The article doesn't mention when such chips would be ready for production and I doubt that IBM's original press release sheds any light on that subject. So all this COULD mean is that IBM only announced their breakthrough ahead of Intel, not that they are ahead or behind Intel.
It's still good to see that Moore's law is hanging in there.
Full Tilt
Does this mean the Phenom will be produced on 22nm scale? Could be a very interesting development in the AMD/Intel chip wars.
Enjoy Every Sandwich
Via their press release: http://www-03.ibm.com/press/us/en/pressrelease/25147.wss
Dual processor motherboard. Problem solved ;)
Try the horse's mouth - there's more detail.
Wyrd, dude.
Using some of SCO's intellectual property, of course...
Do you have ESP?
"...but chipmakers have hit a problem in that current lithographic methods are not adequate for designs as small as 22nm owing to fundamental physical limitations. IBM claims to have solved this problem."
This is virtually the same statement made every time a smaller fabrication process is announced. It conveys no information. Obviously some physical limitation was preventing them from making smaller circuits, and then they overcame them to make them even smaller.
LS
There is a fine line between being a cultivated citizen and being someone else's crop. - A. J. Patrick Liszkie
In other news, scientists have discovered that veggies shrink the brain...
I'd like to see somebody do something new besides just get smaller. CELL for example.
Most users are just fine with a fixed system on a chip with no PCI. (ram too if you could pull that off) If you want to reduce power and cost you'd place as much as possible on a single chip. (using crazy IP games they could buy designs for parts on the chip-- consolidating manufacturing as well.)
How about a working variation of Hyperthreading? have 1.5 CPUs and manage it so almost runs like 2 full CPUs? (since pipelines are still problems.)
At least AMD is going to combine GPUs. But next they need to think about how to better integrate the vector processing that GPUs are taking over - instead of the weak MMX/SSE/etc features which have a lot of overlap in their uses.
How about hardware accelerated stacks? MMUs that can handle a driver memory space (not just kernel and user.)
Advances in clockless processing?
Just slapping more cores on chips is the lazy way out. Most people could use a business-class computer on a single chip with a stick of ram. maybe even a slower cheaper but larger secondary ram...(since GPU ram would get used a lot doing all that fluff that every OS now has.)
Democracy Now! - uncensored, anti-establishment news
Let me translate the press release:
We announce that our future product, someday in the undefined and possibly distant future, will hit 22nm. We're making partnerships to make it happen.
The slashdot writeup is misleading. For shame!
Maybe they did achieve 22, but perhaps there's are tiny catch: They don't work. They only claimed 22nm, not working 22nm. Watching all this Nov.2008 campaign coverage has taught me to read between the lines.
Table-ized A.I.
What a joke of an article. Every semiconductor manufacturer has several generations of process in various states in the lab. Woo IBM's showing sneak peaks at 22nm!
...it sort of takes 3D transistors and all that, but we know how to do these things. It's all using standard silicon, it's CMOS it's extraordinarily well charictarized right? But we've got transistors running at 11 nanometers, I can show you photographs of them. We have the leakage issues but we've got a very good plan.
I met with an Intel VP for an interview a while back and talked about where things are going. He had some nice lab-pr0n of what the photos claimed were 11nm transistors. I believe it was said that was "about 15 years out", and meant to offer reassurance that Moore's Law still had a bit more time left to go.
Actually here, let me go dig up my transcript so I can get a proper quote:
You're going to see that platforms are going to continue to evolve. We're moving to a faster cadence. The processor cadence is about a two year cadence, in terms of process technologies. By the way this is interesting. We know how to do Moore's Law for about another fifteen years which we've never had that kind of length of projection before.
That was 2 years ago, early October 2006. Who leapfrogged what now?
Introducing the new Occam Fusion! Now with sqrt(-1) fewer blades!
Here in Denmark we want our chips big and crunchy. Silly americans' chips are so small they can drink them from a mead-horn.
If you quote this signature there'll be 72 copies of Windows ME waiting for you in Heaven.
It is no secret that IBMs legal department and their patent portfolio is what always gives IBM the upper hand. I'm sure that they managed to get alien cpu technology in an settlement for alien infringement on one or more of IBMs many patents.
Thomas S. Iversen
http://www.eetimes.com/news/semi/showArticle.jhtml?articleID=10810046
Though a more recent article stated that the first plant using 15nm won't be online until late 2011, or early 2012 at the latest.
In the silicon production market there is usually about a 5 year, or more, period between when something is announced, and when it is in production. Which means we will see IBM's 22nm process as early as late 2013.
Microsoft, Apple, Google, Amazon what's the difference? All steal money from devs and control with walled gardens.
The linked page contains a flash ad that took over IE. It injected some chinese ad into other sites that also used flash.
Don't open this page in IE.
They just used Shrinky Dinks and cook up those processors in the toaster oven.
Well, unfortunately it's a bit like the problem with conspiracy theories: anything that needs the complete cooperation of thousands to keep a secret, isn't going to really stay a secret. Building a 22nm fab is going to require a lot of stuff, and a lot of people knowing what is being done there, how, and why. It takes only one disgruntled employee, or some chinese subcontractor going, "hmm, I wonder what'd they buy that big an electron gun for... too big for electron microscopy... could it be they're using electrons at this many electron-volts instead of light?" to lose that trade secret in a jiffy.
A polar bear is a cartesian bear after a coordinate transform.
Indeed. I loved the boxes when they appeared; I could just count the lines when scrolling down and immediately know that x reply was to a troll post about goat vaginas instead of the threadjacker's interesting wall of text. The new/pre-old-new post layout (this) is ugly and disorienting.
This reminds me of how people once said "Microsoft is the largest software company in the world." Well, actually IBM is. *sigh*
What about Power7
So, with this new chip process, I can expect my G5 PowerBook... This Fall?
And make sure we break that 3Ghz barrier. Best not keep Mr. Jobs waiting any longer.
It's mandatory to wash your hands before returning to the land of Dairy Queen.
They computed it down... DUH!
... ... ...
how many pixels do you see?
I'm still in college and we have a big semiconductors lab, so we had to learn the basics of lithography in class. The problem that people are running into is that everything uses UV light, which theoretically can make details of 10nm (its wavelength) but this is incredibly hard. There exists, but not commercially viable, techniques which use x-rays (masking material an issue), electron beams and proton beams(deBroigle wavelength). If IBM got one of these to work commercially it would be a big deal. If they built a state of the art one of these and made some 10nm features, no big deal. Probably the single biggest issue is that they have to make a machine accurate enough to be exactly in the focal point of the beam(~0.1 nm) and the smaller the beam you are using, the smaller the focal point so making more precise machinery is as much of a limiter as small beams.
Stop saying 'deep magic' !
Now people this does not just give a leg up to IBM but also AMD since they have a agreement to share manufacturing tech. What happens if AMD just goes right to 22nm and skips 32nm all together. Gets them right back on par with Intel for process technology. Obligatory wikipedia link
I get a strange magic,
oh, what a strange magic,
oh, it's a strange magic.
Got a strange magic,
got a strange magic.
Silicon on ShrinkyDink.
It uses the same 65nm process, but then shrinks down to 22 in the oven.
A fundamental problem in making smaller features on a chip is light diffraction. At those sizes, the light interferes with itself and what you get on the chip isn't what you put on the mask. (Random site I found seeming to explain it: http://www.pieter-kok.staff.shef.ac.uk/index.php?nav=research&sub=litho)
One way around this is to use smaller and smaller wavelengths of light, but we're now pretty much at the point where that isn't feasible.
Another way is to modify the actual mask pattern to account for diffraction effects, which is what it sounds like IBM just said they're going to do. You make a pattern, calculate how diffraction is going to mess it up, then add bits around your features to change the diffraction pattern. Hopefully, those little extra bits are what get destroyed by diffraction when you actually pattern and the original desirable feature is what's left.
Source: I'm an MSE undergrad
We are not hitting one physical limit. We are hitting several physical limits. Feature size, dielectric effective thickness, poly-Si depletion and band separation are the ones that I know aboot. We are seeing diminishing marginal returns, due to physical limitations.True, shrinking you feature size will give you more devices for the same size die. However, shrinking the feature size also increases power consumption. It does not increase efficiency as the article claims. When the feature size is shrunk, the gate dielectric thickness has to be reduced accordingly. As the dielectric shrinks, leakage (via tunneling) increases. This is why the idle power has increased dramatically over the years. Switching to HfO2 dielectric has helped. There is however, another problem, namely the poly-silicon contacts. A transistor is all about charging and discharging capacitance. The capacitor you want to charge and discharge is the one that's under the gate. All other caps are parasitic. The devices today are switching so fast that the poly-Si is being deplete of carriers. This causes an increase in parasitic capacitance. The next problem will be band splitting. As you reduce the length of the gate, you are using less atoms. As the number of atoms drops, the bands of each individual atom become "visible". It becomes difficult to tune the band gap as the bands split (to which band do you tune?). Hence the multi-core processors. Multi-core is the only way the industry can keep following Moore's rule.
Ever put a new cotton shirt in the dryer on high for about 2 hours too long? They're gonna make 45 nm chips and scale them down by throwing them in a laundry machine.
Something I recall a while back, IBM was working with a large X-ray Laser system for manufacturing very small scale chips instead of light. Problem at the time was there was no way to focus the x-rays and use a proportional size masque.
A google search for "IBM X-ray Lithography" turns up quite a few interesting hits.
Sounds like the technique that TrueType fonts have to resolve the problem of trying to render fonts at low-pixel fonts. If no corrections are performed, the character (or glyphs) will either merge into each other or skip particular segments of glyphs (eg. missing out the middle bar of the letter 'm'. The font-engine actually usually a 'virtual machine' with an instruction set that performs geometric calculations (like project point to line, snap point to grid, set axis of projection line) to solve this problem.
Vintage computer adverts: http://www.vintageadbrowser.com/computers-and-software-ads
The two key technologies are a more advanced form of "source mask optimization" (SMO) than has been used to date, and "ability to program the lithography illumination source at the grayscale pixel level". A very readable account is here: http://www.semiconductor.net/article/CA6597205.html More detailed information on current and older SMO techniques can be found at these links: http://www.luminescent.com/pdf/PMJ08_079_Minimize_MEEF_in_Low_K1_Lithography.pdf http://wps2a.semi.org/cms/groups/public/documents/membersonly/van_schoot_presentation.pdf
As someone who has worked with 130, 65, and 45 kits, I predict that working with 22 as a designer is going to be a huge mess. 65 was the last "normal" generation, where layout actually looks like you would expect. Starting with 45, you need to add lots of "dummy" devices in order to make layout printable. You open the layout for something as simple as an inverter and have a hard time actually finding the transistors. This makes manual layout all but impossible (or at least, take 10 times longer than it used to). Designers will be almost completely constrained to standard cell based designs, which will really hurt in high performance areas.
Did Apple retain the capability to start making PowerPC Macs again, or have they washed their hands of IBM's tech? IBM just might become CPU king once again (stranger things have happened).
That that is is that that that that is not is not.
maybe they'll be out in time to go on a manned mars mission.
Comment removed based on user account deletion
Some secrets are easier to keep than others. It depends less on the context, more on the secret. Of course, you believe every one of those "leaks". Wikipedia is unreliable, whereas leaky secrets in the general environment are irreproachable. Shifting sands of credibility. I suspect you'd have to level up ten tables at Texas Hold'em to achieve your first belt in industrial counter-espionage. Some pretty important secrets have leaked, and some pretty important secrets didn't. Every secret on its own terms.
45/2 = 22.5 round down. Computationaly scaled.