"We designed the Cell processor to be applicable in more applications than just games," says Peter Hofstee, Cell chief scientist and architect for IBM. "Mercury is a superb example of a company that can take technology like this that is different from what is in the market now and start building up support that will enable it to move into a wider market."
Mercury is a developer of embedded computer systems for medical-imaging, defense, and seismic-processing applications, and provides both board-level products as well as supporting software. Under terms of the agreement, Mercury and engineers from IBM's engineering and technology services unit will collaborate to develop products that can be used to handle graphics-intensive workloads and other computationally intensive applications. Initial hardware testing has shown that the Cell processor can provide performance of about 200 billion floating point operations per second.
"Mercury is a company that really understands parallel processing," says Nathan Brookwood, an analyst with Insight 64. "I think Mercury could help create products for applications that might otherwise need to use a whole bunch of floating-point digital-signal processors."
Brookwood says he's skeptical about the use of the Cell processor in more traditional computing environments like PCs or servers because there isn't software that's been specifically tailored to utilize the processor.
The relationship between IBM and Mercury is intended in part to help create those new applications, Hofstee says. "They can help bridge this technology and make it available to a wider set of companies," he says.
IBM has been working in its design centers to enable the Cell processor to be used with a Linux operating system, Hofstee says, but in most "embedded" applications where the Cell is expected to be used initially, it will likely be used with real-time operating-system software.
Rebecca Austen, director of deep computing marketing for IBM, says the company has been working on simulations using the Cell processor within its On-Demand computing centers. The simulations are attempting to determine how a cluster of Cell processors might function compared to more traditional processors in given applications and with a variety of software.
IBM may look to build an array of computing systems based on the Cell processor, she says, and offer computational time to customers on a per-usage basis as a method of creating new applications and systems based on the architecture.
Stay connected and informed by visiting the CA Solutions Center Community!

Become a member today for instant access to free InformationWeek research, expert advice, peer perspectives, and more on the following topics:
- Application Performance Management (APM)
- Security Management
- Mainframe 2.0
- IT Automation
- Service Assurance
Also, visit our Government and Financial Services groups to see how these technologies apply specifically to those industries.
NOTE: Offer valid for U.S., U.S. possessions, & Canada only.