News & Analysis
Comment
snowboard9
C VanDorne
And how about this question: does this technology enable Intel to change the ...
Intel details 22-nm Ivy Bridge at ISSCC
Rick Merritt
11/21/2011 8:00 AM EST
SAN JOSE, Calif. – Intel will describe as many as four designs using its new 22-nm process technology with 3-D transistors, including one of its next-generation Ivy Bridge processors at the International Solid-State Circuits Conference in February.
Intel will discuss an entry-level desktop or notebook processor from its upcoming Ivy Bridge family. It uses four of its IA-32 cores, a graphics-processing core, memory and a PCI Express controller, all built with 22-nm 3-D transistors.
Intel gave its first public disclosure of Ivy Bridge at the Intel Developer Forum in September. The company claims the 22nm process with its tri-gate FinFET transistors delivers twice the performance or half the power compared to its 32 nm process. Four Taiwan ODMs showed prototype ultrabooks using Ivy Bridge chips at IDF.
Three other ISSCC papers will discuss pieces of Intel processors made in its new 22-nm process. The company will describe a 22nm digital phase-locked loop (PLL) and a reconfigurable clock-generation core based on the PLL. The 22-nm core consumes 3mW at 1V and operates at 3.2 GHz in low power mode.
In a separate paper, Intel will talk about a SIMD vector graphics core that uses 22-nm technology to get a nine-fold increase in energy efficiency over its prior parts. Another Intel paper will describe a 32-nm Intel IA-32 processor core that consumes just 737 mW at 1.2V while running at 915MHz.
Researchers from China's Fudan University will show an even more power efficient part at ISSCC. Their 16-core processor made in 65-nm CMOS consumes just 320mW at 1.2V while running at 800 MHz.
At the high-end, Fujitsu will describe the K computer, currently ranked as the world's fastest supercomputer on the Top 500 list. The first version of the system used 548,000 Sparc64 VIIIfx cores and a proprietary Tofu interconnect to hit 8.162 petaflops. An upgraded version with more cores recently became the first system to surpass 10 petaflops.


goafrit
11/21/2011 11:10 AM EST
This 3D thing will surely make me to attend ISCC. Intel is pushing this technology like nothing else. They will surely make AMD to give up this decade.
Sign in to Reply
kytsai
11/21/2011 6:11 PM EST
AMD can go "even fabless" to take advantage of advanced process technologies provided by leading foundaries with competitive manufacturing cost.
Sign in to Reply
wilber_xbox
11/21/2011 12:00 PM EST
any paper on the process information used to fabricate the FinFET and in general? There has been lot of talk about Intel presenting something relevant for the process engineers and folks in this field.
Sign in to Reply
rick.merritt
11/22/2011 1:38 AM EST
I didn't see it in the ISSCC preview on their Web site but there are dozens of papewrs and I may have missed one
Sign in to Reply
walken1
11/21/2011 2:09 PM EST
Intel posted some materials on its 22-nm process on its website but its a few months old and less drilled down for public consumption: http://download.intel.com/newsroom/kits/22nm/pdfs/22nm-Details_Presentation.pdf
Sign in to Reply
elwood
11/21/2011 8:00 PM EST
Without a doubt, FinFet is the next "free lunch" in process technology. Is TSMC or GloFo even close to releasing anything similar. High density/low power SOC will give Intel an edge for sure.
Sign in to Reply
rick.merritt
11/22/2011 1:39 AM EST
GloFo said they don't need 3-D transistors until the next node.
Sign in to Reply
snowboard9
11/24/2011 8:57 AM EST
LOL. Don't need it or don't have it?
Let's go with 'we don't need it' !
Sign in to Reply
Oscar Law
11/21/2011 9:53 PM EST
From planar transistor to 3D transistor, it takes time to learn DFM tricks for new layout style, the other foundries need to spend more efforts to catch up Intel.
Sign in to Reply
resistion
11/22/2011 12:07 AM EST
I'd be interested to read how each fin would be contacted.
Sign in to Reply
C VanDorne
11/22/2011 1:24 PM EST
And how about this question: does this technology enable Intel to change the Boulean paradigm and go base-3? So instead of ones and zeros we'd be talking zeros, ones and twos?
Sign in to Reply