MemTest86 Error Test 5, 7, only multit-hread 8, 32 gb ram 4 sticks

[jasper]

Hi,

I have testet MemTest86 V5.1.0 Free Build: 1000 (64-bit),
My machine,
i7-4770 3.4 GHz
ASUS Z87 Deluxe
Crucial DDR3 Ballistix Sport VLP 32GB (4 x 8GB sticks)

Testet 1-9, rounds 4.
Errors at test 3,4,5, 7 and 8
Only when 4 cores and 8 Hyper-Thread enabled.
And also errors when underclock RAM and CPU.
No errors when 4 cores and NO Hyper-Thread.

Test:

Test 0 [Address test, walking ones, 1 CPU]	2/2 (100%)
Test 1 [Address test, own address, 1 CPU]	2/2 (100%)
Test 2 [Address test, own address]	2/2 (100%)
Test 3 [Moving inversions, ones & zeroes]	1/2 (50%)	4
Test 4 [Moving inversions, 8-bit pattern]	1/2 (50%)	14
Test 5 [Moving inversions, random pattern]	0/2 (0%)	22
Test 6 [Block move, 64-byte blocks]	2/2 (100%)
Test 7 [Moving inversions, 32-bit pattern]	0/2 (0%)	91
Test 8 [Random number sequence]	1/2 (50%)	4
Test 9 [Modulo 20, ones & zeros]	1/1 (100%)

Lowest Error Address	0x7C0CD0F68 (31756MB)
Highest Error Address	0x7C0CD0F68 (31756MB)

Last 10 Errors

[Data Error] Test: 5, CPU: 1, Address: 7C0CD0F68, Expected: 86F91D2E, Actual: 86F91D2A

Now I have only have errors on test 4,5 and 7,8 and always errors with Hyper-Thread.
I have tried to change the RAM in the four places many times.
No problem with only 2 sticks and 16 gb ram.
But the error only changed once.
constant error in addr 0x7c0cc0fe8 and it changed to 0x7c0cd0f68. And now it is always 0x7c0cd0f68. It could be changed that I have tried different settings in BIOS (UEFI).

How do I know what the problem is?
a) Memtest86 v5.1.0 or if
b) RAM problem ?
c) Motherboard problem?
d) CPU problem, memory controller in i7 ?

I really hope it's Memtest86 thats the problem. Because no problem with without Hyper-Thread (seen problem in forumv4.2), and there is no problem with 16 gb ram and the error does NOT change address place.

Hope you can help me.

[keith]

The fact that it is only a single address with only a single-bit error seem to indicate that the error is hardware related.

Do the errors occur in single CPU mode as well when hyperthread is enabled?

Can you also post or upload the MemTest86.log file under EFI\BOOT\.

[David (PassMark)]

The difference between the two error addresses might be the result of you swapping the order of the memory sticks in the motherboard slots. But with dual channel memory and interleaving the addresses it is hard to predict which addresses correspond to which stick. Or as you point out it might be BIOS settings related.

Is the machine stable in normal use, or did you start running memory tests because of some other strange behavior?

You said the error doesn't happen with just two sticks, but which two sticks did you try? Maybe this was with the two good sticks.

[jasper]

The error only occurs when I have hyperthread enabled.
Only error when using all 8 hyperthread CPU with Memtest86.
And no errors when No HyperThread, and only 4 cores (i7-4770).

I could not attach a log file so here a Pastebin for Memlog86:
http://pastebin.com/download.php?i=jWPCmCnj

[jasper]

The difference between the two error addresses might be the result of you swapping the order of the memory sticks in the motherboard slots. But with dual channel memory and interleaving the addresses it is hard to predict which addresses correspond to which stick. Or as you point out it might be BIOS settings related.

I have tried to swap the four memory sticks 6 different ways:
abcd, abdc, cdab, cdba, dcab, cdab

Allways the same memory address, and I now know after change a lot of options in BIOS (UEFI), that the address change back and forth between the single address with error.
Example:
Error in address 0x7c0cc0fe8 and it changed to 0x7c0cd0f68 because of a change in BIOS. So it must be how the computer accress the ram.

Is the machine stable in normal use, or did you start running memory tests because of some other strange behavior?

You said the error doesn't happen with just two sticks, but which two sticks did you try? Maybe this was with the two good sticks.

I have had some unstability with windows 7, even after reinstall that why I tried it.

[David (PassMark)]

Here are the two errors (from your post and the log)

Test: 5, CPU: 1, Address: 7C0CD0F68, Expected: 86F91D2E, Actual: 86F91D2A
Test: 4, CPU: 1, Address: 7C0CD0FE8, Expected: 7F7F7F7F, Actual: 7F7F7F7B

In both cases the error is a single bit (bit #2)

Also the log shows the top address is 0x83FE00000, which is much higher than the address of the error location. I was originally thinking that we might be testing past the end of the address space, but that doesn't seem likely.

Multiple different tests are effected. Making it less likely it is a software fault in MemTest86.

This combined with Windows being unstable make me think it is a real hardware fault.

So the trick now is to work out if it is the CPU, motherboard, RAM slots or the RAM itself which is at fault.

Did you test 2 sticks at a time, but in all of the slots?

[jasper]

This combined with Windows being unstable make me think it is a real hardware fault.

Then I think I should get a hardware shop to look at it. But it's really weird. But I have a hard time believe that it's the RAM.

Did you test 2 sticks at a time, but in all of the slots?

Yes, I have also testet 2 pair of sticks at a time, in the same two slots. This gave no errors.

I have looked at the CPU temperature in Memtest86, at it first starts when the cpu is about 62-65 celcius hot!

[David (PassMark)]

I think I should get a hardware shop to look at it

Unfortunately most of them won't be in a position to do anymore than you have already done in terms of RAM testing.
What they might have however is more spare parts to try out.

65C is hot for an idle CPU. Hot enough that you should check the heat sink and fans.

But 65C at full load is nothing to be concerned about.

Also check for BIOS updates. From time to time there are BIOS bugs that cause areas of RAM to be flagged as free RAM, when they should be reserved RAM. Which can result in RAM errors in MemTest86 and O/S crashes.

[jasper]

65C is hot for an idle CPU. Hot enough that you should check the heat sink and fans.

But 65C at full load is nothing to be concerned about.

But is Passmark using all cpu to ram testing, is the CPU then hot when 65C ?

[David (PassMark)]

The CPU temperature is updated at the start of execution of each test.
CPUs take 1 or 2 minutes to reach maximum temperature under full load.

[jasper]

I have tried Prime95 in Windows 7, but not BurnIn.
I stopped when it reached temperature 80 c. This took only a couple of minutes.

But it only gets up to 66 c in MemTest86 with all CPU.

[David (PassMark)]

MemTest86 wasn't designed to generate the maximum CPU temperature possible, that isn't its job. Due to the nature of the tests the CPU will be idle some of the time waiting for the RAM to respond.