Lessons from Linley

By Ryan Baxter - 2014-02-21

Though I left Boise the morning of February 5 at 0-dark-hundred to catch a puddle-jumper down to SFO, my morning fatigue soon waned as I found myself standing under bright lights at a Hyatt Regency podium a little after 10 a.m. addressing a room full of engineers and system architects.

This was my second appearance at the annual Linley Data Center Conference which was held in Santa Clara, CA February 5-6.  One of several conferences put on by the well-respected research group, this event covers system design for data-center servers and networking, so I am definitely in my element with this crowd.  I was the last of three speakers on a panel entitled Server Technology following presentations by Applied Micro Circuits and Freescale.  My talk, called “Developing Scalable and Resilient Memory Systems with Hybrid Memory Cube,” which you can view below, focused on HMC adoption, comparative technologies today, and the status of the HMC Consortium.

At a high level I touched on what I think are three critical areas for those looking to up-level their memory solution for supercomputing and networking / data centers:

  • HMC is sampling now -- we’ve proven it’s a viable product and folks are actively designing in the technology.
  • HMC provides flexibility and reliability to break into more mainstream computing applications in the areas of acceleration and distributed computing.
  • There is increasing interest building around the HMC technology in the enterprise computing space.

At the conclusion of the session the line of questioning matched the long line at the front of the room. It was the closest I’ve come to feeling like a rock star…sort of. :-) There was a high degree of interest in the Consortium, what it takes to become a member and how companies can contribute long term.  People wanted to know about HMC connectivity to current systems, active efforts to enable HMC in the acceleration space, high performance distributed computing and embedded applications.  I fielded some questions about HMC connecting to areas we didn’t naturally expect, like applications associated with the Rapid IO community.  And there was a lot of talk about current technologies that don’t match the breadth HMC can offer.

All in all, kudos to the Linley Group.  The event drew highly technical, highly intelligent engineers that, while they may not all have been supercomputing guys, were a very engaged, broader tech community to whom Micron was able to expose the promise of HMC.  These were corporate IT decision-makers eager to learn about the newest technologies that can be incorporated into next-generation architectures, and by the looks of the business cards and emails I received after day one, it was time very well spent.  (And the tiramisu wasn’t bad, either.)

Ryan Baxter

Ryan Baxter