Abstract Preview

Here is the abstract you requested from the IMAPS_2012 technical program page. This is the original abstract submitted by the author. Any changes to the technical content of the final manuscript published by IMAPS or the presentation that is given during the event is done by the author, not IMAPS.

3D Integration of Wide IO Memory Cube Stacking to 28 nm Logic Chip with High Density TSV through a Fabless Supplier Chain
Keywords: Wide IO Memory, 3D TSV integration, Fabless
Wide IO memory has higher IO counts (up to 16x) than typical low power DDR memory, which could enable higher system bandwidth at low power. Wide IO DRAM can be stacked as Micro Pillar Grid Array (MPGA) cubes, which will provide high memory density for the system. With the high number (~1200) of connections to the MPGA, a direct face to back stack (3D) to logic chip with high density TSV is the most efficient approach. To utilize the extra large bandwidth, the logic chip containing high speed processors requires logic chip fabrication in advanced node devices. In this paper, we will report the demonstration of a 2-memory chip wide IO stacking on logic chip through a fabless supplier chain. The logic chip is fabricated at a foundry with via middle TSVs in 28 nm node. Middle of line backside processing was performed at the foundry. Thin wafers were shipped from foundry to OSAT and the MPGA is stacked to the backside of logic chip with ubumps at the OSAT. A successful integration of via middle TSV to 28 nm logic process has been demonstrated with minimum impact to logic devices. The final package showed good TSV and ubump integrity through 3D X-rays between logic and memory chip, and within memory chips through 3D X ray inspection. The wide IO memory is functional post stacking. In addition, the early reliability data for TSV and ubump showed no detrimental impact through temperature cycle and high temperature storage. To enable HVM productization of MPGA for mobile wireless applications, a quick learning for high yield is critical. In addition, supply chain business models for stacked memory cubes and competitive pricing are still to be resolved. Initial adoption of this disruptive new technology will depend on offering a compelling technical and business value proposition which cannot be attained by extending existing technical solutions
Sam Gu, Senior Staff Manager
San Diego, CA

  • Amkor
  • ASE
  • Canon
  • EMD Performance Materials
  • Honeywell
  • Indium
  • Kester
  • Kyocera America
  • Master Bond
  • Micro Systems Technologies
  • MRSI
  • Palomar
  • Plexus
  • Promex
  • Qualcomm
  • Quik-Pak
  • Raytheon
  • Specialty Coating Systems