twistedbrain said:
Hi!
I've the following nasty problem: a PC with a motherboard Gateway
GA-G31M-ESL2 rev 1, F8 with 2 slot for memory and 2 DIMMS Corsair
TWIN2X 4096-6400C4DHX (2 GB each). I tried many time to install
different Linux flavours, Kubuntu and OpenSuse, but after some tens
minuts the process freezes and the PC too, I have to push the power
button to restart it. The memory should be fit for the mobo because
Corsair on its website says so, also if the Corsair one could work with
CAS 4-4-4-12 and at 2,1 V, instead the requirements for the DIMMs for
such mobo are CAS 5-5-5-18 1,8 V.
I launched memtest 4.0 that you can find with OpenSuse and after few
minutes it reported some thousands errors in test3 (Moving inversion 8
bit pattern) starting about from 1800 MB where the pattern dfdfdfdf
become dfdfdfdb (errbit 4). But If I test each memory stick in each
memory slot I don't get errors. There are errors only when the memories
work together in dual channell.
Now either mb, either RAM are new and I've the right to the warranty,
but which one is the broken one and why? (I don't have other mb or other
DIMMs to check).
Best regards to all and thanks to replying people,
Andrea
~
Run dual channel, swap DIMMs. Where did the error go ?
Did the errored bit move ?
When a DIMM has a boosted voltage spec, you can interpret
that a couple ways. When I see "CAS 4-4-4-12 at 2.1 V",
that tells me it really needed the voltage, to have
the tight timing. If you loosen up the timing, like your
5-5-5-18, then you may be able to use less voltage. The
industry standard voltage might be 1.8V.
I see some DIMMs, with specs like "1.9 to 2.1V" and I
don't know if that means "timing met at 1.9V" and
"can withstand 2.1V" or what to think.
In any case, you can start by repeating the dual channel
test, and swapping the DIMMs, so that if it was the DIMM,
the errored bit should move. Remember that both the
address reported and the bit location could be important,
because the size of the test location is 32 bits, while
the width of a DIMM is 64 bits. So be really careful
when examining the results, to make sure you're interpreting
it properly. Write down some addresses as well as the
errored bit pattern, before you swap DIMMs. Then do the
math when you run the test with the DIMMs swapped, and
with some luck, you'll be able to tell if the
error moved with the DIMM. In other words, the address
value is also important, in identifying what is bad.
Either the DIMM is defective, or a little extra voltage
will make it work. Up to the stated limit. Some memories
don't tolerate excess voltage well, and a high voltage
may limit the operating life. It wouldn't help, for example,
if the DIMM gets really hot with the excess voltage. Some
people who apply lots of voltage, install a cooler over
top of the DIMMs. And some brands of DIMMs, don't need an
excuse to fail. I check the reviews on Newegg, before
I buy memory now, even if I end up buying the memory
from a Canadian E-tailer.
There is an alternative explanation. There is a difference
in optimization, between the BIOS setup for 1GB DIMMs and
the setup for 2GB DIMMs. Some older motherboard BIOS are
only set up to work well with 1GB DIMMs. The BIOS doesn't
have good settings for a 2GB DIMM. My own motherboard works
that way, and there are no plans for another BIOS release.
It means I can install 2GB DIMMs (and have tried them) but
they throw errors. And I'm not going to waste the rest of
my life experimenting with bus drive settings and other related
crap. If I had a digital storage scope, it might be
worth tuning them, but just randomly trying settings
is for idiots
I'd have something like 256 combinations
to try. It means I continue to use 2x1GB in that system,
and those are unbelievably stable. (More stable than the
previous generation of PC3200 I was using.)
If your motherboard has a BIOS update available, you can try
that. Make sure the motherboard is stable - use your one DIMM
config, before attempting to update the BIOS. Select a stable
method for updating the BIOS. On my system, that is a
DOS boot floppy and flasher program. Examine your update
options, and select the one that is least likely to
freeze on you. Doing it from Windows, is just asking for
trouble. If you have no other choice, but to flash
from Windows, make sure the tool has downloaded the file
to your local disk, so a networking problem won't
cause the flash update to fail.
(I also like to archive the current BIOS file, using the
archive option on the flashing tool. It means, I always
take a copy of the original BIOS, before flashing. If
the flashing tool says it cannot complete the flash, I
immediately try to reflash the original BIOS in. To
increase the odds the board won't be bricked. So have
your archive file in hand, before doing it. Once you
push the reset button, at that point you're gambling
the flash was a success.)
If the memory error stays put, when you swap the DIMMs,
then you might blame the motherboard. Maybe bumping some
voltage can fix it. It depends on what options you have,
as to whether that might work out.
Paul
