Errors when running all tests, but not when I run the same test individually

[MVB]

I recently ran MemTest86 in parallel core mode, and saw that Test 9 was giving me errors (like 1 or 2 errors per pass). I reran the test and saw that the same thing was happening only test 8 gave me errors too. I then tried running tests 9 and 8 individually, choosing only cores that were giving me errors in the previous tests. The results were good with no errors! I then ran the individual tests in parallel mode, and again no errors. And once more in parallel, but this time I had errors.

I ended up updating my BIOS and I'm currently rerunning the tests. Up through pass 3, test 8 has given no errors, but test 9 has given me several.

I'm not sure I understand why the sequence of tests in parallel would give me grief, but the individual tests would be fine in single core mode.

Any thoughts?

[David (PassMark)]

Might be temperature related.

Or it might just be random and not the cause and effect you suspect. (Doing multiple runs at each setting to confirm consistency of results would eliminate this possibility however).

[MVB]

I think you may be right. I just ran a single core run on test 9 and got an error.

Is it common for errors to only show up in one or two tests? I would think if the RAM was bad it would fail all the time...

Also, do errors definitely mean it's an issue with RAM, or could it be CPU or even motherboard related?

[David (PassMark)]

While nothing can be ruled out, there is a high probability it is the RAM at fault.

Some type of RAM error (like a bit stuck high or low) will flag errors on many of the tests. Other errors are more subtle and depend on timing issues or the value of the RAM, or even the values held in adjacent cells. (or temperature, or power supply spikes, or EMI).

[MVB]

After running several more tests (I tried each stick of ram individually, and once I found the RAM to be good, I tried it in different slots,trying different combinations of RAM in various slots, and finally trying different BIOS settings) I found that the only thing that gave me problems was running more that 2 sticks of RAM with XMP turned on. Parallel CPU testing had nothing to do with it, and I imagine that in the end neither did temp (though I'm sure XMP made the chips run hotter, as well as almost no gap between the sticks for airflow). Either way, not sure what to do. Is it worth it to RMA the RAM, or just deal with XMP mode disabled? I ran a couple of tests with MemTest86 and the difference was hardly noticeable.

[David (PassMark)]

XMP runs the RAM faster and hotter. So if the ram is marginal it isn't surprising that you see the faults with XMP on.

You might find a slight voltage bump might improve the XMP stability.