Memtest86+ Odd errors on new build

Abssorb

Distinguished
Oct 13, 2008
26
0
18,530
Perhaps the experts can advise please?

Memtest86+ showing Test#7 fails with millions of errors with a new batch of ram failed. Then, all ram replaced, and replacements failing in the same place. Tested each stick one-by-one, and singly in each slot - all fail the same way:

Pass 42% or 43% Test#7 [Random Number Sequence]
http://abssorb.com/hosted_images/pc/memtest/Order_524256_After_RAM_changed.jpg

It's a new machine which was built for our small business, it's pretty high spec as it's going to run a few virtual machines. The supplier is being really helpful, but TH has more expertise I reckon.


CPU
AMD FX-8350 Eight Core CPU (4.00GHz/8MB CACHE/AM3+)

M/B
ASUS M5A99X EVO R2.0

RAM
32GB SAMSUNG DUAL-DDR3 1333MHz (4 X 8GB) - M378B1G73BH0-CH9

M378B1G73BH0-CH9 is verified as being on the QVL for ASUS M5A99X. The bios/UEFI was upgraded to latest and factory defaults set.

It's seems so unlikely for this many hardware faults with two batches of new ram. Any suggestions welcome. :)
 

Abssorb

Distinguished
Oct 13, 2008
26
0
18,530
Also - forgot to mention:
The machine actually boots and runs OK, (Ubuntu 12.10 64bit) the only symptoms are -

This first started with a strange problem, the PC speaker would give a single constant tone of varying length, from a few seconds to 30 secs+ when the CPU was being pushed. It would also occasionally do this after booting (not post, after during OS boot). While investigation was going on, we tried a suspend in ubuntu (12.10 64bit) which failed badly and so ram testing started.
 

raja@asus

Distinguished
Sep 28, 2011
891
2
19,360
You may need to relax some of the memory timings and subtimings to get this kit fully stable.


I would add +1 to CAS and tRCD to start. Then add +2 to read to read and write to write. Try than on a single module and see what difference (if any) it makes to stability.


Also check the DRAM voltage applied matches the kit spec.
 

Abssorb

Distinguished
Oct 13, 2008
26
0
18,530
Thanks Raja :) My research uncovered a problem with memtest86+ though, so this is solved, but might be of interest to others. I wish I'd found it earlier, I've wasted a lot of time.

The memtest86+ v4.20 which is included on current 12.10 Ubuntu iso downloads is faulty and is giving false positives for test #7.

Other sources of memtest86+ v4.20 do not produce the errors, it's just the one on the Ubuntu download. There is a fix in the repos, but this hasn't made its way onto the isos yet.

https://bugs.launchpad.net/ubuntu/+source/memtest86+/+bug/1071209