Glitch hits ServerWorks chipsets
Broadcom slows shipments of a ServerWorks chipset containing an error that could cause systems to stall. It's already sent out a diagnostic test for computer makers.
The company, which makes communications and networking chips, said that the hardware error is in its Grand Champion LE chipset. It is limited to a small number of systems that use the chipsets, which connect the processor in a server with the memory and other components. The problem affects only those system designs that use two Intra Module Buses (IMBs) at once. The buses are a way for the chips within the chipset to communicate with one another.
All of the computer makers that use the chipset have been notified, Broadcom said, and the chipmaker has developed a software utility to test whether individual motherboards are faulty. The Irvine, Calif.-based company did not disclose the manufacturers affected, but said many of its customers used the product. Dell Computer, Hewlett-Packard and IBM are among those who use ServerWorks chips.
"ServerWorks has initiated a program for its customers to return product for rescreening," said Duane Dickhut, CEO of Broadcom's ServerWorks unit, in a statement. He added that because it takes time to get enough testing equipment in place, the company expects "a temporary delay in (its) ability to meet current demand for this chipset."
A Broadcom representative declined to say how many of the chipsets have shipped or how many have the bug. In addition, the representative did not have an estimate for the cost of replacing defective chips and fixing the problem.
However, Broadcom said it did not expect the issue to have a material effect on its finances.
The company has known about the problem since early June, the representative said. It is not clear whether any makers had yet shipped systems with the chips. Broadcom sent its customers software to test for the problem about three weeks ago, the representative added, although it did not publicly identify the problem until now.
IBM, which uses the chips in its eServer x345 and eServer x245 products, said it identified the problem during testing and alerted Broadcom.
"Since IBM identified the problem early in laboratory cluster-stress testing, our customer exposure is limited," a representative for Big Blue said. "IBM worked with ServerWorks to put an effective screen in place, and we are working with a handful of customers that have experienced the problem."
Hewlett-Packard said it has also been aware of the problem for some time and has been working with Broadcom over the past two months on the issue. The company said it started shipping two dual-IMB models earlier this month--the ProLiant ML370 and the ProLiant DL380--after the screening software became available.
In March, Broadcom dismissed the head of its ServerWorks unit, Raju Vegesna. Earlier that month, the company said it had a manufacturing problem that was slowing shipments of one of its Grand Champion chipsets. A representative said that production issue was unrelated to the current problem.