The PowerPC alliance Thursday responded to online reports of a bug in its new PowerPC 750 processor. The companies said that while the high-performance chip occasionally displayed a floating-point anomaly during tests, the issue will not occur in real-world operations.
Apple and Motorola RISC Microprocessor Division of Austin, Texas, said that during testing some PowerPC 750 processors showed a sensitivity to electrical noise when performing an artificial series of floating-point operations. The noise could cause an error in the floating-point data.
The companies said they encountered the problem only infrequently, and it didn't arise with all the chips tested. "As far as we can tell - after 15 months of testing - there's nothing there," said Phil Schiller, Apple vice president of desktop and server systems.
Apple last week introduced new lines of midrange desktops and a high-performance PowerBook that use the 750 processor, which Apple calls G3. The chip is also offered in several third-party upgrade cards (see stories, Pages 1 and 25).
Schiller said the problem crops up only when the processor is sent "nonsensical opcode sequences," such as repeated floating-point register stores followed immediately by multiple floating-point instructions.
Schiller said this type of event should not arise in real-world computing. He said in qualification tests performed by manufacturers IBM Microelectronics Division of East Fishkill, N.Y., and Motorola, as well as by Apple and third-party vendors, "not a single error was found."
According to Motorola, the issue was treated as a non-critical technical issue and was covered in its online documentation for the processor on the PowerPC home page.