Gimme an "A"....
Gimme an "M"!
Gimme a "D"!
(lover of underdogs)
Netgear on Thursday said that four of its product lines may experience "higher-than-normal failure rates" and that it is contacting affected customers to assess whether replacement or repair is appropriate. "While this situation is not an immediate cause for concern about the operation of any of these models, we understand …
Ahh yes, whereas Intel with the processor serial number scandal, dodgy dealings with Rambus, P3 1.13GHz recall, the whole of the terrible P4 line-up etc. was just doing superbly all along right?
Oh, and the Athlon XP for a sizeable chunk of its lifespan was the fastest x86 CPU available...
Meanwhile Netgate deal with the issue in their very own way by shutting down support forum threads:
Due to the increasing amount of false information spread by individuals with questionable intentions this thread is now locked. We encourage all Netgate and pfSense customers to contact us via official support channels if they have any doubts or questions.
Opening a new thread will result in a 30 day ban."
Very poor form!
Ouf of four "dead" Atom AOA1xx netbook boards, only one was utterly stuffed.
One was the bad BIOS IC problem, one had a defective graphics issue which I repaired
by replacing two capacitors and the third had memory controller problems.
Might have just been a bad socket because it worked with my test 256M stick but not a 512, 1 or 2GB.
The really bad one had low core voltage but no bus activity.
Strangely enough the BIOS chip from dead board *4 worked fine in the bad BIOS IC one.
All had broken sockets which I bypassed with some solder-fu.
Since bought another two more recent duocore units, of these one works fine but has bad controller IC so no network or SD slot, the other *seems* working but again has liquid damage.
Caveat: may not be relevant but still useful. I suspect the problem isn't the clock generator but in fact the output circuitry on die.
I would speculate that its a "capacitative load" issue caused by driving a large copper track up and down with a fast transistor. The inrush current spikes can get very large and cause elecromigration in the power rails. The QA meeting must have been interesting:
"You overloaded the output transistor"
"No we didn't"
"Yes you did ... for 3 Nanoseconds every clock pulse :-)"
Would explain why costs are being shared. Cisco took a $125M hit last quarter on this issue. Intel's share was undisclosed but large.
Why would a microcontroller clock circuit be driving a large track load? I don't know if this Atom device has an external CLKOUT but if it does, surely any sane designer would buffer it before driving anything that looks like a bus?