System instability at long test runs (memtest86+)

[covision]

Dear Toms Hardware forum users.

I am running Memtest86+ v1.70 to test the memory stability on my systems.

Normally, after 100-400 passes in Memtest86+ (2-4 days of running), a memory error occurs on two of the systems.

The memory used on both systems is Kingston DDR2 HyperX PC6400 2048MB CL4 (KHX6400D2LLK2/2G) running at 2.0V (also tried 1.8 and 1.9) with timing 4-4-4-12 as described in the specifications. They are running at 800MHz (set in bios)

Motherboards on the two systems:
- Gigabyte GA-P35-S3G, 333MHz FSB
- Gigabyte GA-EP43-DS3, 333MHz FSB

The only voltage I have adjusted is the DDR RAM voltage (2.0V).

Processors used:
- Intel Dual Core 2.33 GHz, 7 x multiplied

I assume that my system will not be stable in the long run when Memtest86+ fails, even after such many passes.

I would appreciate very much any help to make these systems 100% stable.


with kind regards,
Thomas Evensen

 

[outlw6669]

2-4 days of memtest
Now that is a bit of overkill.
Most people would call it 100% stable after 8 hours of testing, the rest would say that after 24 hours.
Under normal usage, I would say that you would have about a 0% chance of having stability issues because of your RAM.
What are you using your system for that requires you to test the RAM for 4 days, might I ask?

 

[covision]

2-4 days of memtest
Now that is a bit of overkill.
Most people would call it 100% stable after 8 hours of testing, the rest would say that after 24 hours.
Under normal usage, I would say that you would have about a 0% chance of having stability issues because of your RAM.
What are you using your system for that requires you to test the RAM for 4 days, might I ask?

Thank you very much for your response.

The systems are in use for video recording 24/7 365 days per year, so 100% stability is a requirement.

We have had some very bad experience with faulty memory before, so I am a bit paranoid now


with kind regards,
Thomas Evensen

 

[outlw6669]

Just to clarify, after 2-4 days of testing you have only one error or a string of errors?
If it is only one error, I would call it a fluke and would not be to concerned with it.

If it is a string of errors, try testing each stick individually.
It is possible that one of your sticks is a little weaker than the rest and may not be up to the server like strain.
If all the sticks test fine individually, try dropping your timings to 5-5-5-18 OR raising your voltage to 2.1v.
Dropping your timings should insure stability but at a very minor drop in performance.
Raising your voltage will cause your RAM to heat up more but can improve signal stability and probably bring it to stability at rated timings.

One thing you should keep in mind is that stress testing software puts an artificially high load on your components.
Even recording video 27/7, it is very unlikely that you will strain your RAM as hard as memtest does.
As such, I would consider it highly unlikely that you would have even one memory related recording error.
If your RAM really was bad, it would be spitting errors within the first few minutes of testing.