Memory testing and increasing reliability of backups

[Guest Keatah]

Everyone by now probably knows how much I preach about backups and their importance.

 

Since this so important, I've been going though some past repair logs and journals. I found that people who've complained to me about files "not working" or backups not restoring have some sort of memory issue. This is especially the case when all other causes have been ruled out like media, CPU, bug-free software.

 

1- Use Memtest-86. This will test your system's memory far more thoroughly than anything built into the BIOS or whatever O/S you're using. It runs through with all sorts of patterns and is great for checking flipped bits or interfering adjacent cells.

 

2- Use a program that can verify what it just copied. Whether it be a single file or complete image. If you find yourself ever having to do this out of necessity, you better test main memory. You have a fault somewhere in the system.

 

This is open-source (GPL) and has been around forever. Very well respected. Get full version free here -- http://www.memtest86.com -- and burn your own .ISO Let it run overnight. Any single bit error warrants a repair.

 

Good technical read on what Memtest86 does -- http://www.memtest86.com/technical.htm

And if you send them $10.00 they burn the cd 4U. How thoughtful!

[nodles]

There's also Memtest86+

 

I've both as bootable on my USB stick. Though didn't know PassMark maintains Memtest86 now.

[Andavari]

Some computers can't use Memtest86/Memtest86+, for some an alternative like Microsoft Windows Memory Diagnostic may work.

Edited March 2, 2013 by Andavari

[Guest Keatah]

These memory tests are good at ratting-out motherboards with bad capacitors, noisy power supplies, bad power connectors. Those types of fails make so much electrical noise on the rails that only the best of the best designs can isolate and conceal the real source.

[Andavari]

If lucky sometimes pulling the memory out, removing any dust and cleaning the contacts, and then reinstalling it will sometimes work.

[Alan_B]

Another possibility is to move chips around.

 

When 4 GB was added to my existing 4 GB my P.C. failed to boot several times.

It booted after removing the new chips.

It booted with only 2 GB added to first 4 GB

Failed to Boot with the two new chips in new positions.

 

After playing musical chairs with all new and old chips in different sockets it eventually booted with the full 8 GB accommodated by the motherboard.

 

I have had no memory (chip) problems that I am aware of,

and Macrium has always validated the partition image backups that it created on my secondary HDD.

 

I really wish that this topic had NOT been started,

I am now wondering how much safety margin exists in the voltage levels and timing of my existing configuration,

and am now afraid that MemTest86 may destroy my delusions

[Guest Keatah]

I really wish that this topic had NOT been started,

I am now wondering how much safety margin exists in the voltage levels and timing of my existing configuration,

and am now afraid that MemTest86 may destroy my delusions

 

Moving chips around is a good idea and is necessary when you're running up against limits. I've got a system here, too, that only works with the 4 modules in a specific order. Took about 20 swaps to get it rolling.

 

But if you're concerned, run memtest and wiggle the timings and voltages a little in each direction and go with the centerpoint. You're not overclocking anything are you?

 

Furthermore, if there is a latent marginal module, best find out now and replace the sucker..

[Alan_B]

Moving chips around is a good idea and is necessary when you're running up against limits. I've got a system here, too, that only works with the 4 modules in a specific order. Took about 20 swaps to get it rolling.

I find it strangely reassuring that you are confident with a system that failed on the first 19 deployment patterns.

Mine only failed on the first dozen and I had almost given up hope of employing the last 2 GB.

But if you're concerned, run memtest and wiggle the timings and voltages a little in each direction and go with the centerpoint. You're not overclocking anything are you?

No - I do not overclock anything - I am old school.

After a 4 years getting an Honours degree in Electrical Engineering, specialising in Electronics,

my first 12 years were spent designing electronic instrumentation using worst case analysis and calculating Mean Time Before Failure (MTBF) according to a military standard "Blue Book".

All my designs were for continuous operation, often in explosive atmospheres in petrochemical refineries.

The next 30 years were spent designing real-time computer hardware and software for continuous monitoring and reporting of fire and intruder detection systems in shopping centres and nuclear and military installations.

I always assume the worst and endeavour to always have at least two fall-back procedures in place.

 

The last time I used MemTest was more than 10 years ago on a one week old Windows 95 machine that compiled the same source code as my previous machine, but produced a different output binary.

The I.T. department called the supplier who sent a technician who spent a couple of hours testing and reported all was well.

I downloaded MemTest and left it running overnight - it scored a result and the supplier swapped the machine for another one that passed MemTest scrutiny.

 

I do not remember doing any wriggling of voltages and timings.with that old MemTest.

Is this something the new MemTest will do for me, or are you suggesting I visit the BIOS and face all its fears an bogeymen

 

Regards

Alan

[Guest Keatah]

Memtest doesn't alter any settings, never did. You have to man-up and face the BIOS. The only time I overclock & eco-clock systems are for testing purposes. I've found that the centerline of a spec is a good thing.

 

Like if you have CAS4, CAS3, CAS2. And fails at CAS5 and CAS1, just go with CAS 3.

 

In the case of the one system where I swapped memory nearly 20 times - there could be a host of reasons for having had to do so. But I believe that timing variances set up a harmonic resonance or some kind. I never investigated beyond swapping the module order. If it the machine can run memtest and windows memory diagnostics and cpu burn in for several straight days. That's good enough for me. You can't ask for much more.

 

I, too, have training the sciences, theoretical physics, cosmology, electrical engineering. With practical experience in semi-conductor fabrication and electronics, data recovery, and psychology.

[Alan_B]

Thanks

 

At the moment I regard the BIOS with fear and loathing so it will be a little while before I dabble with the BIOS.