Memtest 86 v5.01 False Positive or Serious Problem??

[BhavP]

Mobo Gigabyte Z77-D3H
CPU I5 3570K (Not OC)
RAM CML8GX3M2A1600C9 - 8GB (2x4GB) Corsair DDR3 Vengeance Jet Black Low Profile, PC3-12800 (1600), Non-ECC, CAS 9-9-9-24, XMP, 1.5V
I have started to receive BSOd recently on a custom built PC which is just over 13 months old, I have run memtest86 v5.01 and received errors in test#6, 795 errors on the first module and 495 errors on the second module. Is there something wrong with the memory or should I change something in my system? The system has been running with default settings since the installation.

I have four slots of ram and have run the test individually for each module.

Your assistance is greatly appreciated.

[David (PassMark)]

You double posted. But I have deleted your duplicate post.

From your screen shots I can see that you are using MemTest86+. Which isn't our software. The author of MemTest86+ decided to use the same name as our software, so it is pretty confusing.

If you need support with MemTest86+, then you should contact the author.

You can find the original & best version of MemTest86 here.

[BhavP]

You double posted. But I have deleted your duplicate post.

From your screen shots I can see that you are using MemTest86+. Which isn't our software. The author of MemTest86+ decided to use the same name as our software, so it is pretty confusing.

If you need support with MemTest86+, then you should contact the author.

You can find the original & best version of MemTest86 here.

Thank you for your response.

Please accept my apologies for creating the duplicate post, I think it happened when I refreshed the screen.

I have now downloaded your software and am currently running the tests via the UEFI mode v5.1.0

What is the acceptable numbers of passes i should let it run for before concluding the test?

Also I noticed then when I had run the test via the older V4 BIOS release of MemTest86 it had picked up a lot of errors within the first pass however when running via the UEFI mode it hasn't picked up any so far........Is this normal?

[David (PassMark)]

By default it does 4 passes. Which should be enough to pick up most errors.

V4 and V5 results should be mostly the same most of the time. But there are code differences (like V5 is native 64bit & used SIMD) that can cause differences in the results.

If you get any errors, then post another screen shot.

[BhavP]

By default it does 4 passes. Which should be enough to pick up most errors.

V4 and V5 results should be mostly the same most of the time. But there are code differences (like V5 is native 64bit & used SIMD) that can cause differences in the results.

If you get any errors, then post another screen shot.

Thank you for your response

V5 did not find any errors however V4.3.7 did and I attach the screenshot for the first module....



I shall post a screenshot of the second module shortly......

I look forward to receiving your thoughts.

[BhavP]

No errors found on the second module but the problem still remain with the first module. I have then subsequently tried each module in all four slots of my motherboard to rule out a problem with the mobo.......

the first module had the following errors in slots 2-4



Can I definetly say I have a defective RAM or another problem?

Thanks in advance.

[David (PassMark)]

Yes, it seems likely there is a fault in the 1st module.


We are wondering why it passed the V5 test however. One possibility is that the UEFI BIOS reserved this area of memory for its own use. The lower areas of RAM (around the address range where you had the error, 0x0000C876000 are often reserved for memory mapped I/O, meaning they can't be used for testing or for the operating system). Another, less likely, possibility is that the BIOS in V4 should have reserved this area of RAM, but didn't, and the errors are in fact false positives.

Would it be possible to try the following,
1) In V5, use the "View memory usage" function and take a screen shot of the address ranges in the 0x0000C8xxxxx range.

2) In V5 try testing both sticks at the same time. This should force dual channel mode and force a remapping of the physical RAM into new addresses. Which then might make a fault appear in V5 as well. I assume that in your testing above with V5, you only did 1 stick at a time?

Also did you add any new hardware to this machine recently? As it could be the case that some new hardware is doing memory mapped I/O, but the BIOS doesn't know about it.

[BhavP]

I am beginning to conclude that it was a false positive......unless otherwise advised.

I have re-run the test in on each individual module as well as together in both V4 & v5 and the results have come clean....no errors detected. I have subsequently run the test using Windows Memory Diagnostics (standard) and that has also reported no errors.

When the original issue appeared it said it was a result of a hardware failure however this no longer appears to be the case.

The only change I made prior to re-running the tests was to remove all HDD and cd/dvd drives from system. I also looked at each setting in the bios to see if there was any misconfigurations as the only thing that was different from the default settings was the enabling of Virtualization which I use to run Virtualbox.

Also I noted that numbers for L1-L3 cache are no longer being displayed in V4.....not sure why this has happened.

I did not add any new hardware, however in addition to my original setup I have a cd/dvd drive, x2 2TB 3.5" Seagate Barracuda drives and x1 250gb WD 2.5" Black . My OS is Win 7 Ultimate 64bit.

Before the problem occurred I was trying move my default user folders to another drive as I want to install a 128gb mSata drive onto my mobo, so I wanted to reduce the drive space so that I could migrate the OS over.

I have attached the screenshot for the memory addresses you requested, perhaps there are some answers that are helpful to you.


Any thoughts......

[David (PassMark)]

So I think one issue is solved, even if another one has opened up.

The memory usage dump does indeed show that the memory addresses where the errors were reported are being used by UEFI, for "Boot services data". Which means they won't have got tested in V5. UEFI is a much bigger piece of software/firmware compared to BIOS. So it gobbles up more of the RAM. On my machine a substantial 370MB, out of my 16GB, are used by UEFI. Unfortunately we don't know of any way of moving UEFI around in memory to test underneath it.

Having memory corruption in RAM used by UEFI could cause all kinds of random bad behaviour.

In V4 & BIOS, there is also a reserved RAM, but it isn't nearly as big an area as in UEFI.

You can somewhat compensate for this switching around the order of the sticks in the slots. This will then test more of the reserved RAM.

So the remaining issue now is why have the errors you had in V4 yesterday have disappeared. It is also strange the cache information has gone. Maybe the gold contacts in the RAM slots were dirty and a few re-insertions have cleaned them up?

[BhavP]

So I think one issue is solved, even if another one has opened up. Sure has

So the remaining issue now is why have the errors you had in V4 yesterday have disappeared. It is also strange the cache information has gone. Maybe the gold contacts in the RAM slots were dirty and a few re-insertions have cleaned them up?

Not sure if the contacts were dirty, this could be a possibility as I don't normally take the ram out when cleaning the dust that accumulates inside.

I find it very interesting that a 13 month old PC had a hardware fault that resulted in BsoD and a high number of memory failures and then subsequently testing each module and memory slot individually somehow cleared all the errors and when the tests where re-run and the hardware fault was also magically rectified.

[David (PassMark)]

Something has obviously changed between the non working and working states.

But if it works don't fix it.