Is it mandatory to run multiple passes during one session?

[Soprommat]

Greetings.

I want to test RAM, as I understand it, for sufficient accuracy, several passes are needed (up to 8-10 per stick of RAM). Is it OK if I perform those passes not in a single run but in many runs at different time? Currently there are problems with power grid in place where I live and electricity not presented for 24 hours and I can not let all tests run for whole night.

Thanks in advance.

[David (PassMark)]

A couple of points.
1) As soon as you find a few errors, you know something is wrong. The value in continued testing of a failed stick is low (unless something changes, like voltages or BIOS settings, etc..).
2) The chances of finding an error in the 2nd pass are significantly lower than the first pass. 3rd and 4th have even less of a chance of finding, a so far undiscovered, error. Generally 4 passes should be enough.
3) If you run a large number of passes over many days or weeks, eventually you will see a 'soft error'. Especially for non ECC RAM. This will be confusing as it is hard to tell a real fault from a soft error.

I don't think having a gap in time between passes changes the odds of finding an error.

[Soprommat]

Thank you for detailed explanation.

[dma-adm]

Completely agree.
3-5 passes should be enough for most cases. This will properly heat up the memory and cpu. For critical systems like servers, controllers, etc. or systems sent to remote sites it's good practice to run memtest (and other tests!) for several hours - e.g. leave testing for night.
Completely intact system - cpu, mem, mb, psu should run memtest endlessly without errors for as long as you want - for days or weeks.
If you see errors after day- or night-long testing this is bad sign.