IBM, in an effort to expand the use of its PowerPC processors, is opening up the licensing of the chip to enable third-parties to build products for it. The technology giant also will create a version of the chip that will become its standard network processor.
“Our goal with the PowerPC has been to continue to innovate … but also provide a wider range of applications and support for our customers,” said Lisa Su, director of PowerPC and Emerging Products for IBM Microelectronics, in East Fishkill, N.Y. “Our customers say that they like the PowerPC, but theyd like to be able to use it more.”
IBM has been licensing PowerPC technology to a select few customers, Su said, but now is widening that group. For example, as part of an agreement with IBM, Synopsys Inc. is creating a synthesizable PowerPC core, according to IBM. Designers who license the core from IBM will then be able to use DesignWare IP tools from Synopsys, of Mountain View, Calif., to create products based on the PowerPC core. In the second quarter, Synopsys will give licensees of its DesignWare Library access to models to help them design the synthesized core.
Also, Cadence Design Systems Inc., of San Jose, Calif., will incorporate the PowerPC into its design services. The agreement with Cadence will enable PowerPC-based designs to be manufactured in both IBM and third-party foundries, simplifying the chip design process and bringing 65 nm and 90 nm capabilities to a greater number of products. That should be up and running the second half of the year, Su said.
According to IBM officials, the company has been creating a number of versions of the PowerPC, including the PowerPC 440GX embedded processor and the Power4 and Power4+ chips for IBM servers, as well as versions for game systems such as Nintendo.
The company now will accelerate a planned convergence of the PowerPC and its own network processor roadmaps, Su said. The company planned to make the convergence after the next two generations of the PowerPC, but now will do so after the next generation, within the next 12 to 18 months, Su said.
Network processors tend to need specialized architectures, Su said. By enabling a general-purpose chip act as the network processor as well, enterprises can save money by not having to invest in that specialized architecture, Su said.