sk8v memory problem

[stueken]

Since about a year now my asus sk8v runs with 2x1G Kingston memory
quite stable. Now we decided to add two additional 2G modules,
but the system seems to become quite unstable :-(

My 64Bit Linux kernel crashed after a few seconds. After some
further attempts to boot, the bios started to emit messages like:
SYSTEM FAILED DUE TO CPU OVER-CLOCKING, although I used all default
settings. Finally the system did not boot at all, not even a bios
came up anymore. After i cleared the NVRAM, it came back again.
But I could not detect any strange settings, and I also reset to
all default settings. The bios version is also quite recent:
1003.008 2005/06/22.

I found some suggestions about a possibly weak PSU, but my PSU
powers 450W, so I can't believe its too weak.

With only 2x1Gb it runs quite stable with both pairs individually.

Both pairs are slightly different:

http://www.ec.kingston.com/ecom/configurator/PartsInfo_disc.asp?ktcpartno=KVR333X72RC25/1G

and the new pair:

http://www.ec.kingston.com/ecom/configurator/partsinfo.asp?ktcpartno=KVR333D8R25/1G

Should I try to downgrade some memory timing to get a stable system?
If it runs stable instead I would accept some "underclocking".

I wonder about the DDR voltage. The bios allows 2.6V to 2.8V whereas
the memory module has 2.5V printed on its label. Is this a problem?

any further ideas?

 

[Paul]

Since about a year now my asus sk8v runs with 2x1G Kingston memory
quite stable. Now we decided to add two additional 2G modules,
but the system seems to become quite unstable :-(

My 64Bit Linux kernel crashed after a few seconds. After some
further attempts to boot, the bios started to emit messages like:
SYSTEM FAILED DUE TO CPU OVER-CLOCKING, although I used all default
settings. Finally the system did not boot at all, not even a bios
came up anymore. After i cleared the NVRAM, it came back again.
But I could not detect any strange settings, and I also reset to
all default settings. The bios version is also quite recent:
1003.008 2005/06/22.

I found some suggestions about a possibly weak PSU, but my PSU
powers 450W, so I can't believe its too weak.

With only 2x1Gb it runs quite stable with both pairs individually.

Both pairs are slightly different:

http://www.ec.kingston.com/ecom/configurator/PartsInfo_disc.asp?ktcpartno=KVR333X72RC25/1G

and the new pair:

http://www.ec.kingston.com/ecom/configurator/partsinfo.asp?ktcpartno=KVR333D8R25/1G

Should I try to downgrade some memory timing to get a stable system?
If it runs stable instead I would accept some "underclocking".

I wonder about the DDR voltage. The bios allows 2.6V to 2.8V whereas
the memory module has 2.5V printed on its label. Is this a problem?

any further ideas?

I would try looking for a memory setting like "command rate" and set it
to 2T. The BIOS may offer [Enabled] or [Auto] as options, so I suppose
[Enabled] would give you 2T (the setting is not shown in the manual).
(I'm not sure whether that should really be necessary. You already are
paying the "command rate" penalty by using registered DIMMs. The register
unloads the address bus, and that should make it easier to drive the
DIMMs. So, while you can try adjusting command rate, in theory it
should not be necessary.)

You should test the memory using memtest86+ from memtest.org. I don't know
if memtest has any issues with the config you are running or not.

The memory are different. One module uses x4 chips (stacked configuration)
and if your processor supports chipkill ECC, which corrects 4 bit symbol
errors, the chipkill would be a good match for the x4 orientation of the
chips. The new module, with the x8 chips, if an 8 bit value was corrupted,
chipkill would not be able to correct it, and the error would be detected
as uncorrectable. I think that is why the Kingston page is making the
distinction between what is different between the old (discontinued)
x4 module and the new one using x8 chips. If you did select chipkill ECC,
the end result would be more uncorrectable errors indicated, whereas using
ordinary ECC would be able to correct some of the errors.

Your pairs of modules should not be mixed together, and will work best in
A1&B1 for one pair and A2&B2 for the other pair.

Check memory timing settings with CPUZ ( www.cpuid.com ) or perhaps with
a64tweaker. Maybe then you can see what your command rate is doing.
Some a64tweaker downloads are here, plus the thread that tracks development.

http://www.akiba-pc.com/download.php?view.40 (v0.5)
http://home.exetel.com.au/codered/A64Tweaker/A64Tweaker_V0.6beta.zip (v0.6)
http://www.xtremesystems.org/forums/showthread.php?s=&threadid=37345

I have seen some posts from people running other Asus Athlon64 boards,
where mixing two different pairs of DIMMs resulted in problems. But
in that case, the second pair of DIMMs went missing (undetected).
Perhaps your problems are BIOS related, and maybe a chat with
Asus tech support by phone, will uncover the details.

HTH,
Paul