Intel is getting more focused on the challenges in the supercomputing market and has established a new business unit, called Intel Federal, to pursue the contracts with the US government that fund the majority of cutting-edge supercomputing research and development in the country.
In the wake of its acquisition of networking chip maker Fulcrum Microsystems in July, Intel merged its wired networking, cloud computing, and data center server units into a new entity called the Datacenter and Connected Systems Group.
In addition to making general-purpose server platforms, this unit is responsible for crafty hyperscale cloudy platforms, the Many Independent Core (MIC) x64-based coprocessors that are analogous to GPU co-processors in terms of being used as offload engines, and various supercomputer engineering efforts.
But the Feds – and particularly the US Department of Energy nuke labs, where the big bucks are spent in HPC systems, but also sometimes the Defense Advanced Research Projects Agency – like a little more hand-holding and often customized systems to support specific functions. And so Intel has created the new Intel Federal unit underneath Kirk Skaugen, general manager of the Data Center Group, to do that.
"An exaflop supercomputer's performance is the equivalent of every person on Earth making about 150 million calculations per second," a statement from Skaugen about the creation of the Intel Federal unit reads. (This is meant to give us an idea about what powerful computers can do, and perhaps humanize what they do, but let's face it, such statements really don't and IT execs should stop doing it.)
"We look forward to collaborating more closely with the U.S. government on future supercomputing challenges. The creation of Intel Federal demonstrates the strategic importance of these programs and will give us the ability to establish and maintain the unique processes, procedures and controls needed to develop and manage programs with the government."
Adding big HPC brains
Intel Federal will be headed up by Dave Patterson, who reports to Skaugen and who was previously CEO of Optelecom-NKF (now part of Siqura and a maker of network video equipment) and before that CEO of Siemens Government Services.
Earlier this year, Mark Seager, former head of HPC systems at Lawrence Livermore National Laboratory, was tapped by Intel to be CTO for the chipmaker's HPC Ecosystems group.
Seager was in charge of the Advanced Simulation and Computing Initiative (ASCI) program at LLNL, and was responsible for ASCI Blue Pacific, ASCI White, ASCI Purple, and BlueGene/L, which were all based on IBM's Power servers and proprietary interconnects, as well as several plain-vanilla x86 clusters including the Project Hyperion testbed cluster at LLNL.
John Gustafson, an HPC researcher, has been made a director at Intel Research to focus on exascale issues. His resume includes CTO at the floating point accelerator maker ClearSpeed Technology and a stint at Sun Labs when the former Unix server maker landed its $47m DARPA HPCS supercomputing contract.
Intel was among a handful of companies awarded $76.6m by DARPA to cram a petaflops-scale supercomputer into a standard server rack and only draw 57 kilowatts to power and cool the machine.
This is called the ExtremeScale challenge by DARPA; and Cray, Silicon Graphics, and Nvidia are also up for the challenge. Intel is leading one ExtremeScale team with SGI helping, while Nvidia is leading the other team with assistance from SGI.
Intel's Fed unit will have offices near its chip and system labs in Oregon, in its California headquarters, and in Washington, DC.
Over time, Intel Federal will expand out from its HPC mandate with the DOE and take on work from other agencies, including work other than HPC systems and very likely cloudy infrastructure. ®