Intel has discovered a glitch involving some of the I/O controllers in the companys new “Grantsdale” chip set, which can cause a PC to fail to boot up or to freeze.
Between 100,000 and 200,000 chip sets are affected, according to one Wall Street analyst, who published a report for his clients on Friday. Intel had polled its customers for the lot numbers of the affected chips, which were sent to manufacturers before the chip sets official launch on Monday. Samples of the lots are being sent to Intel for further testing.
Intels 915 and 925 chip-set families, formerly known as “Grantsdale” and “Alderwood,” consist of several chips, including memory controllers and the I/O Controller Hub, known as the ICH-6.
The I/O hub controls some of Grantsdales new features, including the circuitry to turn a PC using the chip set into a wireless access point, a capability Intel has delayed. Intel executives have called the chip set “the most significant platform in 12 years.”
The glitch is somewhat complicated. Normally, when a chip is being fabricated, an insulating layer of film is deposited to electrically isolate the chip. That film is normally removed from the die pad area, where the chip interfaces with the pins that connect it to the outside world.
In this case, the thin film on a pad area was only partially removed, causing the real-time clock circuitry to be susceptible to excessive leakage, according to Howard High, a spokesman for Santa Clara, Calif.-based Intel.
“Almost none of it is in the end-user community,” he said of the affected chip sets. All of the PCs that are shipping now are free from the glitch, he said.
Intels High said that Intel was not generally recalling the chip. “Maybe its just semantics, but a recall would be having a bunch of product in the marketplace and asking for it all back,” he said.
“I cant tell you the extent of the problem at this stage,” High said Thursday afternoon. “Id lie if I said I was able to size it, if it was 10 percent of the units. I just dont know.”
Andrew Root, an analyst for Goldman Sachs Global Investment Research in New York, reported that only a small number of chip sets were affected, a number that was confirmed by an industry source. “As we understand it, the problem had (past tense because now appears fixed) to do with current leakage in the chipsets [real-time-clock] feature, and was encountered in certain printed circuit board designs,” Root wrote in a note to clients Friday. “Intels own−brand motherboards did not appear to have this issue.
“The product bug has apparently caused Intel to recall a portion (approx 25 percent) of the several hundred thousand chipsets shipped thus far,” Root added. “This suggests a total magnitude for the recall of 100,000− 200,000 chipsets, or less than $8 million of product on a Q2 revenue base of $8.1 billion.”
The problem may be in the early supplies of the chip set that were shipped to PC OEMs and motherboard makers. OEMs, at this point, are playing it cautious. “Its possible well have to delay our launch by a week or so while we get this straightened out,” said a source at a major PC OEM.
But there is one telltale sign that a PC is affected: a persistent refusal to boot. When asked if the glitch might have affected any early reviews of the chip set, High chuckled. “Its hard to say,” he said. “Reviewers get pretty early units. If a unit fails to turn on, we hear about it pretty quickly.”
Word of the problem surfaced Thursday on HardOCP.com, an enthusiast Web site.
Editors Note: This story was updated with additional comments from Intels High and Goldmans Andrew Root.