SGI Computing Platform Puts Petaflop of Power in a Server Cabinet

At the International Supercomputing 2010 show, SGI announced a new computing platform that takes advantage of general-purpose GPUs from Nvidia and AMD's ATI business to scale up to a petaflop of computing power in a single server cabinet.

SGI has created a computing platform that takes advantage of graphics processing capabilities to get a petaflop of compute power into a single cabinet.

SGI announced the hybrid computing platform June 1 at the International Supercomputing 2010 show in Hamburg, Germany.

Like a growing number of server makers, SGI is taking advantage of the capabilities in Nvidia's GPUs for use in general-purpose workloads in the HPC (high-performance computing) space.

According to SGI, the new computing platform offers GPU processing not only through Nvidia's Tesla technology, but also from AMD's ATI graphics business. It also incorporates accelerator-based technology from Tilera, a company that offers 36- and 64-core processors and is promising 100-core chips.

The new platform will leverage all the technologies to get a petaflop-or a thousand trillion floating point calculations per second-into a single server cabinet. SGI officials say the computing platform is aimed at science and engineering businesses that use high-end software.

"Our innovative technology will enable users to attain new levels of scalability and speed with groundbreaking performance capabilities, culminating in a petaflop in a cabinet," Eng Lim Goh, senior vice president and CTO at SGI, said in a statement.

The new technology will begin appearing in SGI products by the end of 2010.

Also at the show, SGI officials announced that they have completed the first shipments of the high-end Altix UV 1000 servers. The HPC systems, first introduced in November 2009, can scale up to 2,048 cores and can support up to 16 terabytes of shared memory.

"The Altix UV platform is transformative to how customers can deploy the next generation of computing for the world's most demanding work loads, including traditional HPC, databases, I/O and cyber security," SGI CEO Mark Barrenechea said in a statement.