SGI unveils integrated blade platform
- By Rutrell Yasin
- Jun 27, 2007
Silicon Graphics yesterday unveiled the first of a new line of blade servers built to handle high-performance computing applications and large-scale-out workloads, company officials said.
The company is targeting the users of clustered servers who want better performance and reliability out of their systems, said Alison Ryan, vice president of business development at SGI.
The SGI Altix ICE 8200 addresses the major challenges these users face such as system manageability, performance, scalability, power, cooling and facilities issues, she said.
Altix ICE combines several commodity technologies such as InfiniBand networking functionality, Intel Xeon processors and the Linux operating system with SGI's high-performance computing capabilities.
The 8200 incorporates a next-generation processor board called Atoka, co-designed by SGI and Intel for high-performance computing, said SGI chief executive officer Robert Ewald at a press event in Washington. The board lets a single SGI Altix ICE 8200 blade be powered by two Dual or Quad core Intel Xeon processors and offers up to 32 gigabytes of memory.
In fact, a single SGI Altix ICE 8200 rack can be powered by as many as 512 Intel Xeon processors cores and deliver 6 teraflops of performance, SGI officials said. The SGI Altix 8200 is available in configurations ranging from 8 to 512 processors per rack. A full rack of 512 Xeon processor cores is approximately $350,000.
'It is absolutely cable-free in terms of the rack itself,' Ryan said. 'That is significant for a rack that holds 512 cores,' she said. The cable-free rack unit and InfiniBand network interface cards built on the Atoka board help reduce the number of failure points in each rack, according to SGI.
SGI faces competition from IBM and Hewlett-Packard, which have already established a presence in the high-performance arena with blade-based systems, according to a recent IDC report on Altix ICE. However, SGI's system could be an attractive alternative for users with medium-sized to large-scale-out workloads, the report states.
Rutrell Yasin is is a freelance technology writer for GCN.