Weird RAM issue or Ubuntu at fault ?

[a7ilcy]

Hello folks, I faced a strange problem last night.

I just set the timings, frequency & voltage of my RAM as per the company-recommended settings at 1333MHz(using memory ratio equal to 5 on my motherboard), 9-9-9-24 @ 1.5V for the particular model.

The thing is I tested the settings in memtest86+ for over an hour without errors (one complete pass).

However, when I booted into the OS, and started doing some web browsing, the system rebooted. This happened twice.

I am still using the same settings, and nothing has happened yet.

My questions:

1. Could this possibly be a software issue ?
2. What could have led to the reboots ?
3. I have been running the system now for over two hours now, and nothing has happened yet. Should I be worried still ?


System Specs: Core i7 920 D0 | MSI X58 Pro-E | Corsair TR3X6G1333C9 6GB Kit | WD Caviar Blue 500GB * 2 | MSI N9500GT-MD1G/D2 | Corsair VX450W | Thermalright Venomous X | CM 690 | Ubuntu 9.10 64-bit

[Kano]

First try a BIOS update, if it does not help return ram and get something officially supported with 6 memory sticks:

http://www.msi.com/uploads/test_report/TR10_1804.pdf

[a7ilcy]

Thanks for the reply.

There are 3 RAM sticks, 2GB each.

BIOS is the latest version.

Well over 6 hours now, and no issue.

Do you still suggest RMA ?

[Ex-Cyber]

It could still be a RAM problem (memtest86 isn't 100% conclusive), but sudden reboots with no error messages are often caused by power faults (e.g. marginal PSU or mainboard power components). I wouldn't rule out ACPI-related problems either.

Consider trying some kind of load testing (preferably not loading the GPU; large compiles or archiving can be pretty effective). If you're using the Nvidia blob, repeat the tests without it as well.

Also consider trying to reproduce the problem on a different OS; if not Windows, then at least something like Damn Small Linux that uses sufficiently different versions that there aren't likely to be lots of shared bugs.

[Kano]

For 3 modules only s1366 is maybe overkill, 4 modules with 2 gb each on s1366 may be even cheaper - basically only for 6 modules s1366 is needed.

[a7ilcy]

It could still be a RAM problem (memtest86 isn't 100% conclusive), but sudden reboots with no error messages are often caused by power faults (e.g. marginal PSU or mainboard power components). I wouldn't rule out ACPI-related problems either.

Consider trying some kind of load testing (preferably not loading the GPU; large compiles or archiving can be pretty effective). If you're using the Nvidia blob, repeat the tests without it as well.

Also consider trying to reproduce the problem on a different OS; if not Windows, then at least something like Damn Small Linux that uses sufficiently different versions that there aren't likely to be lots of shared bugs.

Thanks for the suggestion.

I think I might know the issue. Since turbo-boost was enabled, it is possible that an increase in the base clock frequency caused the RAM frequency to increase...but since it was already at its limit, therefore crossing the limit might have caused a reboot ?

Of course, this is true if indeed my thinking is right. Point of apprehension: does enabling turbo-boost and when it actually works, increases the RAM frequency in real time also ? Or once the RAM frequency in BIOS is fixed, can it not be altered by the controller ?

[Kano]

Turbo Boost changes the internal multiplier not the external frequency. It should not affect ram at all. But if you get lockups with other OS too then it is clear that it is a hardware problem. Some problems only show under load, so memtest is not 100% accurate. It is more like: when memtest fails, then ram is bad, if it does not fail it does not have to be good.

[a7ilcy]

Ah thanks. In that case, will try another distro and see what happens.