3D Thursday: Can IBM and 3M really build a 3D, 100-chip stairway to heaven with glue?

IBM and 3M have announced a program to develop heat-dissipating adhesives that would permit the construction of tall, 3D “towers” of silicon chips in 3D assemblies. The stated target: 100 stacked chips. Some things about 3D assembly are absolutely uncontroversial:

  • Wire-bonded 3D assembly is in widespread use today, pervasively in mobile phone handsets.
  • 3D assembly is clearly the best way to improve performance by minimizing chip-to-chip interconnect parasitic.
  • 3D assembly enables power reduction by allowing smaller I/O swings and slower interconnect speeds using massively parallel interconnect as in the Wide I/O memory interface spec.

However, other 3D issues are controversial:

  • How do you test a 3D stack? Pre-bond? Partial-bond? Post-bond? All three?
  • What interconnect standards are needed to enable widespread use of 3D assembly?
  • How do you get the power out of a 3D stack?

This last question is especially thorny, given the industry’s strong preference for pushing individual die designs to the limits of power consumption. PC processor die routinely dissipate 100W and server die run at almost double that power. Even application processor chips destined for mobile phone handsets dissipate Watts of power.

However, if there were two companies likely to lick the power problem, I’d pick IBM and 3M. IBM has been the world leader in stretching and straining silicon to do it’s bidding since the 1960s and 3M, well what company knows glue—er, sorry, specialty adhesives for the electronics industry—the way 3M does?

Here’s a cool 38-second video from IBM and 3M that shows the concept:

About sleibson2

EDA360 Evangelist and Marketing Director at Cadence Design Systems (blog at https://eda360insider.wordpress.com/)
This entry was posted in 3D, EDA360, Packaging, Silicon Realization, SoC Realization and tagged , , . Bookmark the permalink.

1 Response to 3D Thursday: Can IBM and 3M really build a 3D, 100-chip stairway to heaven with glue?

  1. George Storm says:

    2.5D… is something the world expected IBM would be addressing. So it’s public. Their most important objectives for such work are habitually camouflaged.
    So: one hundred layers? Perhaps for specific applications where individual dice can mostly be de-powered (huge fast memories using NVRAM??).
    If history is anything to go by, they are probably developing more revolutionary technologies in private.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s