Memtest86+ Odd Behavior

[briandm81]

I'm in the process of burning in my FreeNAS box. I'm getting what seems like odd behavior. The first pass ran great and took somewhere around 24 hours. I shutdown after that and did 24 hours of cpustress. That went fine so I shutdown and went back to memtest86+. This time I ran another pass which lasted probably 18 hours before a power outage took it down. So I restarted again and the next pass ran somewhere around 24 hours. So overall, so far so good. No errors, no issues. Here's where it has gotten odd...I'm now 58 hours in and the second pass is still going. So 34 hours after the first pass ran (which took 24 hours) and I'm only at 42%. Should I just restart and let it do one last pass? I've always just used memtest86, is there another test I should run given this behavior? Obviously for FreeNAS, memory is kind of important...

System specs:
SuperChassis 846TQ-R900B
(2) E5-2670 @ 2.6 GHz
Supermicro X9DR7-LN4F-JBOD
256GB Samsung Registered ECC DDR3 RAM (16 x 16GB)
Noctua i4 Heatsinks

Screenshots:
Almost done...

First pass done...second pass seeming slow:

Second pass is STILL going:


Ok...so now I'm at 59 hours and 47%. It's moving...just slower than previous passes. CPU temp looks good, so I'm not sure what the deal is...

 

[gbeirn]

I think you'd be ok. I'm not entirely sure how well memtest actually works with large amounts of system RAM. That and 5.01 has been around for a long time so I don't know if anymore updates are being developed for it. You could ask on their forums.

I'd make sure that ECC is turned on in your BIOS though because I'm pretty sure memtest does detect and report that like ECC + Scrub enabled or something like that and I don't see it in your screenshots.

 

[briandm81]

Thanks for the response. I don't see the specific setting for ECC. I do see scrub settings:

 

[Blinky 42]

Ok...so now I'm at 59 hours and 47%. It's moving...just slower than previous passes. CPU temp looks good, so I'm not sure what the deal is...

It completed a full pass and is 47% done of pass #2.

 

[sth]

it makes sense to run it with all cores engaged which will speed things up drastically as well as load the system maximally. You do that through the initial options page that is presented. Also, get a UPS before you run this box with data, power outages can be bad for ZFS systems. I recall 4 passes on my 256G box took about 24 hours-ish.

 

[Peanuthead]

If you run it with all cores engaged then you may receive false positives.

 

[sth]

Would be curious why? By false positives, do you mean an error reported when there wasn't an error? Or errors not reported?

 

[Peanuthead]

Correct. That is just what I have read.

 

[PigLover]

it makes sense to run it with all cores engaged which will speed things up drastically as well as load the system maximally. You do that through the initial options page that is presented. Also, get a UPS before you run this box with data, power outages can be bad for ZFS systems. I recall 4 passes on my 256G box took about 24 hours-ish.

My experience running Memtest86+ multi-threaded has been decidedly mixed. It was developed for the original "core 2" architecture and I don't think they've stayed current with hardware changes since. I also don't think it is particularly well tested. It is unstable, crashes, and definitely reports false positives. Works sometimes - but when it detects faults you really can't be sure if you have faulty hardware of if Memtest86+ is just being cranky.

Stick with single-threaded. Yes - it takes longer on large memory footprints. But IMO the multi=threaded mode is not stable enough to trust.

 

[briandm81]

I'll have it on a UPS once I get it all set up. It finally finished up the pass...took way longer than the other passes, so that odd, but no errors. Now if I can just figure out why FreeNAS hates my Sandisk 16GB Fit's....I'll be in business.