IBM has started selling the first computer based on its multi-core Cell processor, aimed at companies that run intensive tasks like medical imaging or oil exploration.
The Cell chip was co-developed with Sony and Toshiba and originally designed for video-game consoles. IBM is offering its version of the chip, called the Cell Broadband Engine, in a blade system for customers who want high performance computing power combined with fast graphics performance.
Called the BladeCenter QS20, the new system is also aimed at companies in the digital animation, aerospace, defense, communications and gas industries. Pricing for each QS20 blade starts at $18,995.
The Cell processor consists of an IBM PowerPC processor core combined with eight specialised vector processors. IBM says the processor can deliver "supercomputer-like performance", although most existing applications will likely have to be rewritten to take full advantage of the Cell's architecture.
The company demonstrated a cluster of prototype Cell blade servers at the Cebit trade show in Germany this year. One of the applications on show allowed visitors to view a 3D model of a beating heart, and to turn it around and cut it into cross-sections using a mouse.
Researchers at Germany's Fraunhofer Institute said they took a few months to port the application to the Cell system, since the program had to be divided into many independent, parallel tasks.
The base configuration of the QS20 includes two cell processors each running at 2.3GHz, 1GB of memory, a 40GB hard drive, dual Gigabit Ethernet ports and 1 or 2 Infiniband adapters connected with PCI-Express. IBM suggests using Red Hat's Fedora 5 Linux OS.
Test versions of the blade server have been running at various organisations, including the University of Manchester, IBM said. The university said it picked the Cell system because it offers faster performance while occupying less space and using less power.
In a hospital setting, the BladeCenter QS20 could reduce from minutes to seconds the amount of time it takes for doctors to compare and map 3D medical images, according to IBM. In aerospace and defense it can make radars more sensitive and more accurate, it said.