May 07, 2008
A day after SGI said NASA would be installing a 245 teraflop Altix ICE machine at Ames Research Center, the space agency announced it would be teaming with SGI and Intel for their next generation petascale supercomputer, called Pleiades. That first installation is scheduled for deployment in 2009 and will achieve one peak petaflop. That system will be expanded to 10 petaflops by 2012. Which basically means NASA is committed to keeping up with the Joneses -- in this case, the Department of Energy labs and national supercomputing centers.
No details have been released on the system architecture for the upcoming supercomputer. Ashley Vance at the Register has allegedly been informed that the new Altix ICE machine being built for NASA will "serve as the basis" for Pleiades. Not sure what that could mean other than SGI is planning on an aggressive upgrade path for the ICE platform for at least four more years.
The space agency's big gamble may be sticking with SGI for the long term. As I wrote yesterday, SGI's latest quarterly report showed a net loss of nearly $41 million dollars, on revenue of $79 million. And even though the company has kept operating expenses in the $50-60 million range over the last five quarters, over the past year quarterly revenue has been steadily dropping ($111M -> $78M), while quarterly losses have been increasing ($20M -> $41M). As I write this, the company stock is sitting below $9 per share.
The real bright spot is backlogged orders, which have nearly doubled since Q3 FY07, and now stand at $202 million. The U.S. government also has a vested interest in keeping the tier 1 HPC vendors solvent, so SGI may see some NSF or DOE funding sent its way.
NASA is certainly hoping that SGI manages a soft landing. I'm told that organizational loyalty to SGI runs pretty deep at the agency, and the rocket scientists, especially, love the global memory, Itanium-based Columbia system.
Which brings us back to the question of system architecture for the Pleiades super. It's difficult to envision a 10 petaflop machine based on a scaled up version of either Itanium- or Xeon-based SGI Altix machines. Even considering a Moore's Law progression in processor and memory technology, a 10 petaflop system would suck up at least tens of megawatts. Since Intel has to solve the same problem in its petascale collaboration with Cray, I would assume we'll be looking at a many-core IA processor for a 2012 SGI machine. I'll try to find out more about the Pleiades project over the next week or so and report back.
Posted by Michael Feldman - May 06, 2008 @ 9:00 PM, Pacific Daylight Time
Michael Feldman is the editor of HPCwire.
No Recent Blog Comments
10/30/2013 | Cray, DDN, Mellanox, NetApp, ScaleMP, Supermicro, Xyratex | Creating data is easy… the challenge is getting it to the right place to make use of it. This paper discusses fresh solutions that can directly increase I/O efficiency, and the applications of these solutions to current, and new technology infrastructures.
10/01/2013 | IBM | A new trend is developing in the HPC space that is also affecting enterprise computing productivity with the arrival of “ultra-dense” hyper-scale servers.
Ken Claffey, SVP and General Manager at Xyratex, presents ClusterStor at the Vendor Showdown at ISC13 in Leipzig, Germany.
Join HPCwire Editor Nicole Hemsoth and Dr. David Bader from Georgia Tech as they take center stage on opening night at Atlanta's first Big Data Kick Off Week, filmed in front of a live audience. Nicole and David look at the evolution of HPC, today's big data challenges, discuss real world solutions, and reveal their predictions. Exactly what does the future holds for HPC?