This is part tip, part plea for advice.

I have had a couple of RAM sticks that passed testing with memtest+ for over 24 hours, at about 100 test cycles, IIRC, yet intermittent hardware failures did not go away until I swapped the RAM sticks with different ones. I know that some combination of motherboard + RAM is bad because replacing the RAM makes the problems cease.

The tip part: If you can't track down the cause of bizarre lockups, compilation errors, etc. try replacing your RAM, even if memtest+ says it's ok.

The plea part: Does anyone know of a way to test RAM more thoroughly than memtest+? Right now, I have a set of RAM that always causes the compiler to error out during the compilation of MythTV, but there's no guarantee that this test will catch all the bad RAM that memtest+ does not. It's also not a consistent test - it usually errors out fast, but it's not always in the same place or with the same error.

_________________
KnoppMyth R5.5
MythiC Dragon v2.0
Join the KnoppMyth Frappr!

I don't have an answer to your plea, but I have a vague thought. Some RAM is more forgiving on voltage than others. Depending on speeds, # channels, etc. it can really effect the stability.

When you run a RAMTEST, I presume that you are not really loading the power supply down with processing, encoding, decoding, and reading/writing to the hard drive, so on and so forth. This is really a stretch, but is it possible that your power supply is getting noisy when you are using your system normally, but when you are doing a memtest it is clean?

I'm not normally one to point to the power supply, but I could see it being a problem. If your new RAM is more forgiving than the old, or if you are now using single channel as opposed to dual channel then that can have some bearing.

Not to deviate from your subject, but a situation I had is as follows:
I built a Desktop PC in February with 2 GB of dual channel RAM. On the first boot, everything ran fine, but the timings were wrong. I went into BIOS, set them properly and the computer would no longer POST. Then when I reset BIOS it would power up again. Come to find out, the voltage to the RAM needed to be bumped up. I would imagine that a scenario similar to yours would occur if the voltage to the RAM dropped during operation.

Do you have another power supply laying around that you could test with? Did you ever try operating the unit with only one stick of faulty RAM?

I agree with knappste, though add a different spin.
I'd look at the volts myself as the culprit. I believe memory can be prone to not reporting the correct voltage to the mobo, (or the other way around). However, I solved an instability I had by upping the volts.
I checked with the manufacturer's specs and found they volts being supplied to the ram was down. I set the volts by hand in the bios and the pc has been stable for a couple of years now. I'm using it now infact
(edited to correct grammar)

_________________
Version:R8
Intel C2D 7400, Nvidia 5600 via HDMI to Samsung B37B650TW (PAL), Asus P5QL-E mobo, 4Gb PC6400 DDR2 ram, Samsung Spinpoint 500 Gb & 1Tb drive, Nova-HD-S2 (x2)

Hmm, interesting.

Correct. This was run from the KnoppMyth CD's boot prompt.

It's a 400W PSU from which 90W is drawn when idle and 125W is drawn during peak usage. Neither comes close to the maximum continuous output of the PSU. This problem has happened on two different computers built with the same parts list.

With the realization that low Voltage could be a factor, this becomes even more of a chicken and egg problem. Was this RAM bad, or is the other RAM just more tolerant of bad Voltage?

I could try that, but I still need a test that predicts RAM-related failures, regardless of whether the PSU or the RAM itself is at fault.

No, but I will. I have some RAM that, as pairs, were known to be bad, but I have yet to try them in different combinations. It seems unlikely that they're all bad.

The main problem I have is consistently identifying the bad RAM other than watching for intermittent failures that may only happen over long periods of time. I'd like a process (like memtest+ is meant to provide) where I test it and let it run for 24 hours, and then know whether or not I have bad RAM (or RAM that doesn't get along with the Voltage it's getting).

_________________
KnoppMyth R5.5
MythiC Dragon v2.0
Join the KnoppMyth Frappr!

That's a great insight, tophee. My original problem still remains, though. I still lack a test that will automatically tell me whether the system is prone to RAM-related failures, whether they are due to Voltage, bad RAM, bad motherboard, bad CPU, etc.

I guess I can try the compilation test on the other suspicious RAM and see if it fails. It's a little more cumbersome than memtest+ but may work better. And if I compile with the right options (to use both cores), it'll do a CPU heat test, too.

_________________
KnoppMyth R5.5
MythiC Dragon v2.0
Join the KnoppMyth Frappr!

What are the specs of the hardware you are having problems with?
I was going to ask over at another forum I use. I'm fairly sure they could come up with some answers to your questions, though I think if you are trying to resolve a specific issue with a specific hardware combination they'd I'm sure they'd want some specifics. (Too many specifics?)
Ta.

_________________
Version:R8
Intel C2D 7400, Nvidia 5600 via HDMI to Samsung B37B650TW (PAL), Asus P5QL-E mobo, 4Gb PC6400 DDR2 ram, Samsung Spinpoint 500 Gb & 1Tb drive, Nova-HD-S2 (x2)

The problem hardware is specific to a subset of a batch of RAM I bought several months ago. I just need a test that can tell in advance if a given RAM stick is going to cause failures, which is what memtest+ is supposed to do. Such a test should be hardware-independent, right?

_________________
KnoppMyth R5.5
MythiC Dragon v2.0
Join the KnoppMyth Frappr!

That's why I prefaced my message with

I do not know of such a hardware test. Memtest seems to be the standard, and if there was something better out there I would have thought we would know. This is where computers become more of an art than a science, I'm afraid. It would be a great tool to have though... there is typical no way to tell what causes a lockup, it could easily be RAM, motherboard, processor, or power supply related, among other things. If you do find something, I would be interested to know what.

A number of years ago when I was working in a computer retail workshop we had a hardware RAM tester. What you would do is plug the DIMM into the socket and push a button and it would run a series of tests on the RAM. After a about a minute your would have your result if the DIMM was good or bad.

For the life of me I can't remember the brand of the device, but I can tell you that they do exist and my boss at the time was very explicit that it cost over $10000 (AU) and touted the fact around town that we were the only place in town to have one.

Just did a google and found this product: http://www.memorytesters.com/sim2/s2ltplus.htm and this is the device that we had. Not sure if exact model, but that was it.

_________________
Girkers

I appreciate your help, knappster. I understand that you were offering alternatives for how I could solve the problem, and I'll definitely let everyone know if I find another method of RAM testing.

_________________
KnoppMyth R5.5
MythiC Dragon v2.0
Join the KnoppMyth Frappr!

Yeah, I had considered a dedicated hardware RAM tester, but at a cost of between $400 and $1600, it's a lot of money to spend on something that isn't guaranteed to find the defect

Once I've re-tested the bad RAM pairs I have with the MythTV compilation test, I will see if it's actually able to find all the defective sticks. It's more cumbersome than memtest+, but if it finds defects that memtest+ doesn't, it's worth doing.

_________________
KnoppMyth R5.5
MythiC Dragon v2.0
Join the KnoppMyth Frappr!