3-D Chip Design Challenges

Feb 21, 2010 by John Messina weblog
3D-IC with TSVs and inter-layer cooling channels that's enclosed in a sealed housing.

(PhysOrg.com) -- With the increase functionality of electronic gadgets, the need to pack more transistors into a single package is becoming a needed necessity. There’s a big demand, in the consumer market, for smaller packages and in order to meet this demand, 3-D chip technology must be perfected.

By using 3-D design technology, it’s possible to pack more on a single chip as traditional 2-D scaling slows down. There is however design challenges that need to be overcome before 3-D chips become a reality.

At the IEEE International Solid-State Circuit Conference, this February, Imec engineers presented some key design challenges of 3-D chips made by stacking layers of silicon dies using vertical copper interconnects called TSVs (through-silicon vias). Imec is one of Europe’s largest independent research centers in nano-electronics and nano-technology. They employ over 1,650 people from all over the world to work at ’s campus.

By using TSVs larger bandwidths between memory and logic can exist. Traditional wire interconnects limits bandwidth and consumes more space. This limits the processing speed and consumes power while increasing latency; this in turn creates bandwidth bottlenecking. Because TSVs are typically placed 200 micrometers apart, bandwidth bottlenecking can be greatly reduce.

Heat is another problem for TSVs because the thermal characteristics are three times more critical in 3-D structures as compared to 2-D ones. The poorly conductive adhesives, that hold the thin wafers together, don’t evenly spread the heat through the chip thereby causing hotspots. This can cause reliability issues in the chip and data corruption.

TSVs (through-silicon vias) connect multiple silicon dies stacked vertically in a single chip. TSVs are spaced 10 micrometers apart, insulated by thin layers of silicon dioxide. Credit: Imec

Mechanical stress is also a concerning factor because the copper TSVs contract faster when they are cooling than their silicon counterpart. Too much stress can hinder transistor performance.

The placement of TSVs to transistors on the chip can change chip parameters such as threshold voltage and drive current. Designers will need to compensate for this by creating blank areas or gaps on the chip that would have no devices.

3-D chip designers will have to consider all these factors and carefully balance the cost of the modified architecture that TSVs require in order to have any significant benefits.

Explore further: New oscillator for low-power implantable transcievers

More information: www.prism.gatech.edu/~rgoel3/GTCAD/ICCAD.pdf

Related Stories

Production begins for Xbox 360 chips

Oct 26, 2005

The IBM chip that will power Microsoft's Xbox 360 console is in production and on target for the device's pre-Christmas launch next month.

Memristor chip could lead to faster, cheaper computers

Mar 17, 2009

(PhysOrg.com) -- The memristor is a computer component that offers both memory and logic functions in one simple package. It has the potential to transform the semiconductor industry, enabling smaller, faster, cheaper chips ...

Recommended for you

Microsoft beefs up security protection in Windows 10

3 hours ago

What Microsoft users in business care deeply about—-a system architecture that supports efforts to get their work done efficiently; a work-centric menu to quickly access projects rather than weather readings ...

US official: Auto safety agency under review

16 hours ago

Transportation officials are reviewing the "safety culture" of the U.S. agency that oversees auto recalls, a senior Obama administration official said Friday. The National Highway Traffic Safety Administration has been criticized ...

Out-of-patience investors sell off Amazon

16 hours ago

Amazon has long acted like an ideal customer on its own website: a freewheeling big spender with no worries about balancing a checkbook. Investors confident in founder and CEO Jeff Bezos' invest-and-expand ...

Ebola.com domain sold for big payout

16 hours ago

The owners of the website Ebola.com have scored a big payday with the outbreak of the epidemic, selling the domain for more than $200,000 in cash and stock.

Hacker gets prison for cyberattack stealing $9.4M

20 hours ago

An Estonian man who pleaded guilty to orchestrating a 2008 cyberattack on a credit card processing company that enabled hackers to steal $9.4 million has been sentenced to 11 years in prison by a federal judge in Atlanta.

Magic Leap moves beyond older lines of VR

21 hours ago

Two messages from Magic Leap: Most of us know that a world with dragons and unicorns, elves and fairies is just a better world. The other message: Technology can be mindboggingly awesome. When the two ...

User comments : 4

Adjust slider to filter visible comments by rank

Display comments: newest first

5 / 5 (2) Feb 19, 2010
With the increase functionality of electronic gadgets, the need to pack more transistors into a single package is becoming a needed necessity.

Is need a needed necessity, or is that a needlessly unnecessary triple-entendre? Inquiring minds want to know...
not rated yet Feb 19, 2010
I recall several years ago IBM showcased a concept of a 3d chip that had tiny cooling channels running through the cube itself to overcome heating by increased volume to surface area ratio.

I believe they used the high-K dielectric material that seperates/isolates the chips electrical lanes as the bedrock to tunnel these nanofluidic coolingpores into the cube.

One disadvantage however, that due to the complex wiring, it was very hard to design and produce (multiple vapor deposits) an effective and efficient nanofluidic internal cooling in the remaining High-K mortar and very expensive at that. So their patent is sitting idle, waiting for oppertunity to come along, such as bottom up nanofabrication getting much cheaper and or 3d chip cubes with more regular repeatetive wiring, such a MRAM.
not rated yet Feb 19, 2010
As we get below 22 nanometres even high K dielectrics seperators get so thin they start to leak and the high k is not thick enough to tunnel your internal nanofluid cooling channels. Ideally one would want to use pure vacume instead of high K material (Sillicon On Nothing) or a liquid High K material that doubles as cooling agent, but those both pose the problem of how to physically support the active 3d chips wiring structure.

Perhaps an answer could come from nanotubes/grapheen as it could both provide lanes with superperb electron conductivity as well is physically very strong at the nanoscale, allowing to build selfsupporting wired scaffold with vacume between the beams, the excellent electron and heat conductivity and heatresistence of the carbon would allow to operate without additional coolant.
not rated yet Feb 21, 2010
A "needed necessity"?
Tell me when a necessity is not needed!
Could we have some English here instead of primary school errors?
What's more, "the need to... is becoming a needed necessity" ? !!