Russia's Moscow State University has a hunger for supercomputers. In the last four years it's contracted with high-performance computing (HPC) company T-Platforms to build two clusters: a 60-teraflop machine called Chebyshev completed in 2008, and the much-faster 1.3-petaflop "Lomonosov" (pictured) that just received an upgrade earlier this year. Finding itself still short on flops, MSU has contracted with T-Platforms yet again to build a 10-petaflop cluster scheduled to be operational in 2013. That's a lot of number-crunching power. Currently, the fastest supercomputer in the world is Japan's K Computer at 10.51 petaflops, but the Department of Energy has contracted with Cray and Nvidia to built one pushing 20 petaflops by the end of 2012.
According to The Register, T-Platforms is planning a mixture of different node types to achieve double-digit petaflop performance. The company will reportedly build the nodes from Sandy Bridge or Ivy Bridge Xeon processors and Nvdia's next-generation Kepler GPU coprocessors. Intel's Many Integrated Core (MIC) architecture could also be included if it's available at design time in 2012. All that silicon will go into a custom water-cooled high-density rack-and-blade setup that makes our "killer gaming PC" look like a pocket calculator.