Re: Straight-Up Replacement For P4P800-E?
- From: Paul <nospam@xxxxxxxxxx>
- Date: Sat, 01 Nov 2008 13:49:32 -0400
marcos wrote:
"Paul" <nospam@xxxxxxxxxx> wrote in message news:gegme6$sud$1@xxxxxxxxxxxI'm a bit curious about your old board. I've always wondered whether
there is a "hidden warranty" for ICH5/ICH5R failures. Perhaps you
could contact Asus Tech Support, and see what kind of response you
get. Before phoning, you'd want your bill of sale and your
box serial number in front of you. The regular warranty is 3 years,
and I'm curious whether anyone in the industry feels any guilt about
the "combusting ICH5/ICH5R". I've read one report of an Asus board
being repaired under warranty, and they actually soldered a new
chip to the board.
The serial number on my P4C800-E Deluxe is 41xxxxxxxx (= 2004, January),
and my date of purchase is May 17, 2004. Mine is still working, but
if the ICH5R does decide to blow, I'd be outside the 3 year warranty.
Paul, is there a known issue with ICH5 failures? If so, it sure hasn't
surfaced with any of my IC7-G's. Or are you saying Asus' use of
passive cooling on their motherboards has more to do with it?
It is not "known", as in being generally acknowledged by all the manufacturers.
Gigabyte is the only company to provide a FAQ page (someone else found this
for me). And while ICH4 is mentioned in this article, the evidence in the
news groups, is that ICH5 is far and away the worst chip.
http://tw2005.giga-byte.com/Motherboard/Support/FAQ/FAQ_456.htm
The evidence is there for you to examine. The burn mark on the chip,
is right over top of where the USB I/O pads get their power. (I looked
up the ball-out diagram for the chip, and overlaid the picture on
the ball pattern.) The failure is a "latchup type", and the I/O pads
electrically form an SCR (silicon controlled rectifier), a PNPN junction,
which shorts the VDD and VSS at the pad. This causes a couple of amperes of
current to be drawn, causing severe local heating. Even if a heatsink
was placed on top of the chip, it makes no difference to the underlying
mechanism.
Damage can go two ways. If the bond wires or other interconnect, burn out
quickly enough, all that is lost, is all USB functioning. The
Device Manager still shows all USB entries (because the logic blocks
are undamaged). In that case, just the USB I/O pad itself has lost
power, and cannot talk to a USB device any more. People report
that none of their USB interfaces work (all six fail at the same time),
but otherwise the computer is unaffected.
If the bond wires remain intact, the area of the fault starts to heat
up, and irreversible damage takes place. Machines with the burn mark,
no longer boot up.
The issue is, why did it happen in the first place ? I found an Intel
technical journal article, where they make the claim that their
I/O pads have a 6KV resilience to electrostatic discharge, which
is high when compared to a lot of other chips. So if the Intel
design was so bulletproof, why would ESD be causing these latchup
failures ? Many other motherboard designs, with chips from Nvidia,
ATI, VIA, SIS, and so on, don't seem to have this problem.
A second potential cause, might be if the 5V compatible interfaces
on the chip, really aren't 5V compatible. In that case, there may
be a sensitivity to sequencing of power supply rails at startup.
So far, I've only heard of a couple people having chip damage
at boot. The vast majority die while the machine is running,
which is more consistent with some other cause, like ESD.
Passive cooling is perfectly appropriate for that chip, because
its power dissipation and normal operating temperature, make cooling
via the chip top, and via the ball array on the bottom, good enough.
They don't install heatsinks, purely as a means to remove the
heat from an ESD failure. If a heatsink was present, all it
would do, is protect the plastic.
I've experienced latchup failures in the lab, but none of mine
were the major "burn" type failures. There is also something
called "micro latchup", where individual CMOS gates can latch
up, and that causes the chip to malfunction rather than burn.
In that case, the PNPN junction formed, is too weak to cause
a thermal incident.
The above is my own analysis of all of the symptoms provided
by posters. None of the above has been obtained from any
authoritative source, so these theories are my own. The Gigabyte
FAQ points at a potential source of the problem, without laying
blame at Intel's feet in so many words.
Whether you experience a problem or not, may depend on how
much you use the USB ports. One poster lost a total of
four or five motherboards to this problem, and I really
would have liked to understand what the guy was doing to
the boards, to make so many failures show up in one
household. For that one guy at least, he seemed to know
the recipe for disaster.
If you have an ICH5 on the motherboard, I'd at least recommend
that you stop using the USB port on the front panel of the
computer. The rear ports would be marginally safer. And using
a separate PCI USB2 card for USB devices, could extend the life
of the motherboard considerably.
Even the USB ports on PCI cards have been known to fail -
just ask the people who have NEC USB chips on a PCI card,
as to how many ports still work. But when a NEC port fails,
nothing burns.
Paul
.
- Follow-Ups:
- References:
- Straight-Up Replacement For P4P800-E?
- From: (PeteCresswell)
- Re: Straight-Up Replacement For P4P800-E?
- From: Paul
- Re: Straight-Up Replacement For P4P800-E?
- From: marcos
- Straight-Up Replacement For P4P800-E?
- Prev by Date: Re: Power Consumption Dual Core?
- Next by Date: Re: OT: Power Consumption Dual Core?
- Previous by thread: Re: Straight-Up Replacement For P4P800-E?
- Next by thread: Re: Straight-Up Replacement For P4P800-E?
- Index(es):
Relevant Pages
|
