Link This | Email this | Blog This | Comments (0)

SEMATECH arrives in Albany with 3D Integration program...contd
October 15, 2007

continuing with coverage of the SEMATECH 3D conference in Albany last week......

Ruchir Puri, of IBM  Yorktown, gave his interesting perspectives on  design and CAD challenges. A key point was that increased resistance of the interconnect lines, due to shrinking of the Cu cross sections, are causing need for additional repeaters (buffers)

 

           i.e. 130nm - 2000, 90nm-20K, 65nm-193K, 45nm ~2-3M

 

 That’s right 2-3 million repeaters for a processor designed at 45 nm. Such buffers cost lots of power and account for > 50% of logic leakage.

  

Juan Xie -  Penn State,  Jason Cong – UCLA and Paul Franzon –NC State discussed design issues, many of which were beyond my grasp, but they left me feeling like things were progressing and under control

 

Muhannad Bakir of GaTech, working with Jim Meindl reported on their program looking at  “fluidic I/O” for heat removal by liquid cooling. Maybe it’s just me but micro channel cooling which has been around for 15+ years does not look like the 3D thermal solution for me. Maybe some day for servers....maybe. 

 

The last presentation was a very intriguing one from Nextreme, a thermoelectric cooling (TEC) solutions company. I have personally been working with them to see if TEC has anything to offer for cooling 3D stacks (that was my full disclosure statement). Seri Lee, ex Intel thermal expert, is their CTO and he described how Nextreme has developed a “thermal copper pillar bump (t-CPB)” which looks like a thermal copper pillar bump except it is sandwiched around a thin slice ( 10 – 20 um) of their  special thermoelectric material (see figure below). When it receives current from the redistribution trace below it, it can pump heat from a hot spot above it into the redistribution trace underneath it and down into the substrate. They claim to have worked with Intel to cool a hot spot by 14 °C (although not using the TEC in this new configuration). The plan is to work with 3D companies to try to pull some of the heat down through the bumps, thus taking some of the burden off the normal backside heat sinks. This thermal CPB is certainly intriguing and I look forward to seeing how it functions in a real 3D stack situitation.

Any questions about this meeting ?? leave me a  message. 

 

Posted by Philip Garrou on October 15, 2007 | Comments (0)