IBM, Samsung, Globalfoundries, and more looking to beat Intel to 28nm market

Sure, Intel's one-upping AMD in the 32nm department, but IBM and its merry band of Technology Alliance members -- including Samsung, STMicroelectronics, and AMD chipmakers Globalfoundries -- are looking to ramp up the competition and develop even smaller, low power 28nm processors before Intel gets a chance to size down. The group additionally promises migration plans for companies who've got 32nm on their roadmap and want to maybe shrink a few of the later, already planned models. Early risk production for the 28nm chips are planned for second half 2010, which means it's very unlikely we'll be seeing them in consumer gadgets until at least 2011.

[graphic contraste @ Apr 17th 2009 5:59AM]

""Early risk production""

[AuDioFreaK39 @ Apr 17th 2009 12:59PM]

32nm and 28nm are not very far apart you guys.

Remember in the past? We went from:

90nm to 65nm
65nm to 45nm
45nm to 32nm

but 32nm to 28nm isn't much of an optical difference.

[maveric101 @ Apr 17th 2009 10:12PM]

dude 90 nm was only like 3 years ago. it wasn't too far in the past that Intel was using a 300 nm process.

[Geekbuzz @ Apr 17th 2009 6:09AM]

Wow 28nm, that's really incredible... juste remember 10years ago

[Engrudget @ Apr 17th 2009 6:11AM]

Can't wait for the next big step--the 27nm.

[Shinigami @ Apr 17th 2009 5:54PM]

Intel will beat all other companies by reaching 26.5nm first!
Not that it matters...

[maveric101 @ Apr 17th 2009 10:15PM]

32 nm to 28 nm is a much bigger difference than you think. it's actually a 24% decrease in cross-sectional area.

[Rebellion881 @ Apr 17th 2009 6:20AM]

I don't see what's the big deal since Intel is slated to go 22nm by 2012 when they release their 8 core variant CPU.

[Geir E @ Apr 17th 2009 6:21AM]

To add to how tiny 28nm is.. The original pentium 60/66mhz chips was built using a 800 nm production process. The grandfather 8086 processor from 1979 was built using 3,000 nm. So in 30 years we are down from 3,000 nm to 28nm. amazing!

[elBravo @ Apr 17th 2009 9:36AM]

3um!! wow! you can practically feel 3um by touch!

[DR House @ Apr 17th 2009 6:35AM]

We already traveled to space million year ago yet we cant create 5NM chip yet? Humans......

[Brent @ Apr 17th 2009 11:33AM]

People with spaceships were around a million years ago? Sweet!

[Kwikit @ Apr 17th 2009 6:54AM]

The genius behind the general purpose processor with all those logic components is absolutely amazing, regardless of size... but when you look at a processor and see the city they built on a square centimeter, it's mind boggling. It's funny... you can't even see 28nm with an unaided eye, can you? The strand is probably so fragile.

[hempey @ Apr 17th 2009 7:26AM]

You definitely can't. 28nm is roughly 2000 times thinner than an average human hair.

[Kwikit @ Apr 17th 2009 1:45PM]

Thanks Hempey, I just confirmed... a human hair is said to be between 17 and 181 um, and if you take 100um/2000 that would equal 50nm. :o That is freekin thin. :)

[Leindurstit @ Apr 17th 2009 7:52AM]

With Intel's jingling bag of gold around their neck and super-power market position, it won't matter much if another company (or even a band of companies) manage to one-up them.

[vdex34 @ Apr 17th 2009 8:24AM]

I disagree, since now the other companies are adopting a wolf pack mentality towards Intel trying to bring it down.

[dg @ Apr 17th 2009 12:37PM]

It won't work. Most of the companies trying to "bring them down" in some way or another buy their products and make money off of them. Whether it's their parent company, or a subsidiary (or whatever the word is for a company owned by another, hah).

[Digital1 @ Apr 17th 2009 8:54AM]

If IBM can wrangle people onto their instruction-sets and get more widespread acceptance I can see them getting close to beating Intel down a bit. IBM has a big presence in things like set-tops,some servers(not as big as x86), and Game Consoles(All modern ones have IBM processors); In these markets IBM will make a lot of gains and probably consume some of the smaller guys. If they hope to overtake Intel they will probably need to pull some delightful tricks... Either way though beating the size race could mean that maybe they can get a foothold in Netbooks running Linux. That would actually be a great application for this I think. Low heat+Low Power+Great ISA=great little computer for surfing the web,etc.

[JW @ Apr 17th 2009 8:59AM]

Rebellion is correct, Intel will likely have sweet yields at 22nm before these yahoos produce anything salable at 28nm.

[Dave @ Apr 17th 2009 10:17AM]

That pic reminds me of a Parcheesi board. Not the same, I know, but that's just how my memory works.

[unspellable @ Apr 17th 2009 11:03AM]

You're not alone: it tricked me twice.

It seems like it would make for an interesting board game or an outline for a city map in an FPS.

[SJ @ Apr 17th 2009 11:10AM]

@Kwikit - 28 nm is roughly 1/20th the wavelength of green light. So, no, you can't see it with the unaided eye. You can't see it with the most powerful possible light microscope, because it is smaller than the wavelength of light you can see. It can only be resolved with an electron microscope.

[Kwikit @ Apr 17th 2009 1:48PM]

Thanks SJ, that really places it in some perspective... I had a feeling it was small enough to require an electron microsope, but I'm just amazed at the fabrication methods... making something that small... but I guess this is gross when you consider that scientists have engraved letters on surfaces only electrons (or was it atoms) wide. :o

[Skorpius @ Apr 17th 2009 11:34AM]

While this is a good thing, how about making them faster, rather than just combining cores? We've been stuck at 3.2ghz stock for too long now.

[dg @ Apr 17th 2009 12:40PM]

You obviously don't understand how this works...

[kal326 @ Apr 17th 2009 12:49PM]

PPC chips have been higher than that for a while, they just aren't your run of the mill x86 desktop chips. IBM Power6 chips go to 5Ghz
http://www-03.ibm.com/systems/power/hardware/index.html

Also some Intel Xeon and Extreme chips got at 3.8Ghz and 3.73Ghz respectively.
http://processorfinder.intel.com/details.aspx?sSpec=SL8P2
http://processorfinder.intel.com/details.aspx?sSpec=SL9AN

[sam @ Apr 17th 2009 6:49PM]

And in performance (a) per clock, (b) per watt, and (c) overall, Intel Core 2 Duo is reportedly faster in client usage. [With Power6 having some advantages at server workload.]

Which sort of explains why everyone's now concentrating on performance per clock and per watt, rather than increasing the clock speed at any cost.

[dg @ Apr 17th 2009 12:31PM]

"...which means it's very unlikely we'll be seeing them in consumer gadgets until at least 2011."

...At which time Intel will be gearing up to go 22nm.

This is pointless.

[gutiea @ Apr 17th 2009 7:52PM]

You must all consider that IBM's Press Releases, never short of breakthrough announcements that go nowhere in the marketplace, has three main audiences: (1) IBM's Board Members and Execs from well performing business units, (2) Legislators in Saxony, Seoul, NY, and Singapore and (3) Demoralized Employees of its R&D and Hardware Technology Divisions. IBM Micro, and most of the partners have been bleeding red for most of the last decade. The partners gave up on trying competing one on one with Intel and decided to all gang up in Public -Private partnerships in which the taxpayers of the above mentioned localities underwrite over 30% of the costs.

Except for Samsung, none of them can make money in this industry anymore, and the 28 nm node is pretty much the last viable business half node. It is not clear that Intel will even try to commercialize 22 nm. The business decision is still pending, although R&D has started in full force now that the 32 nm ramp is about to start.

IBM strategy is not a business strategy, it is instead a survival strategy, a final battle being fought using a risk management strategy instead a product success strategy. Essentially, all the partners conceded that the profits that Intel generates allow Intel to beat them to the ground anytime. Since IBM has about $1.2B (for about 1200 people) in revenue from licensing and less than $4B (and about 6000 people) from product; the business for IBM is next to an annoyance (out of over $120B revenue, and 300k people)

IBM makes these announcements regularly to try to buy credibility, primarily among its own management team and also among the business and government partners.

The situation is getting desperate for all the "ganged partners", particularly because the only company turning profits in this downturn is Intel, and is also the only one likely to emerge with a solid balance sheet.

Keep in mind that business battles are won not only through cool technology but also through good business execution. Intel is superb at both.

This announcement does not even challenge Intel dominance in MPUs; instead, it is mostly a move against TSMC that has already moved to 28 nm for "mostly on-state applications" such as smart phones and the like. TSMC, correctly so, several months ago, identified 28 nm for low power as the key battle ground for mobile. IBM 32 nm HKMG process is almost 30% more expensive per wafer than that of TSMC, although IBM's is potentially faster, a superfluous factor since nobody cares about speed anymore.

The announcement is aimed at assuring IBM partners and remaining customers that they intend to compete with TSMC. IBM and its allies are in fact correcting course midway, to fight two battles: Intel at 32 and 22 nm for computer chips; and....TSMC at 28 nm for smart mobile processors.

I might be wrong, but I suspect that IBM and its allies are headed for their own Battle of Jena somewhere around 2010 and 2011; and that IBM, just like the Prussians, is unlikely to remain in the semi industry after 2013. In fact, I read in this press release that IBM has set the final date for their exit from microelectronics, on or around 2013. Basically, IBM is assuming that the outcome of this final battle will not be that of Waterloo, in which the allies of the “7th coalition” finally defeated Napoleon in the field of Waterloo. (and the Prussians weren’t there)

I guess, I believe the only sure thing is that by 2015 we will all be using Intel chip; simply because, in business war, the Gods always help those that deliver consistent profits.

Just some brief thoughts!