Old version (4.3.7) shows no error, new version does (7.5) what could be the problem?

[merczellevente]

Hi!

I have the following build up:
CPU: i7-3830K @ 3.2ghz
MB: AsRock x79 Professional Fatality p2.20
RAM: Corsair DDR3 quadkit (4x8gb) CMD GX3m4a213 XMD 2132 MHz (with auto latency: 9-11-11-31, and auto voltage: 1.65)

When ran the MemTest86 7.5 Free that consistently delivers errors in test 3, 4 and 5. I did a complete test with zero errors in other tests, then I have tested only these 3.
I have check with each and every memory stick, each delivered error.
The latest results show (I have waited about 30 seconds for the sake of logging the first errors) address errors during test 3: A0133ffc and a10daffc (with 4 errors).
When I have changed the memory sticks to analyze only one, then changed DIMMs, the addresses did not show any comprehensible pattern. Sometimes the same address popped up, sometimes different ones.

BUT! When I have ran the old version 4.3.7 for about 20 hours with the original setup there were no errors at all...
This is critical because I bought the system (cpu, mb and ram as well) second hand (with significant discount btw), so I have to tell the guy whether I want to deliver it back or not.

What do you think how is it possible that the older version delivered no errors? What can be the problem?
I wanted to download previous version (e.g 6 or 7.3) but was not able to find anywhere. Can you please make available for me?

p.s.: i played a little and used some internet but there was no blue screen yet.

An early response would be highly appreciated.

BEst regards,
Levente

[David (PassMark)]

Do you have the error logs?
It would be good to see all the addresses that the error and the number of bits in error.

If all the errors were clustered around the address 0xA0133ffc and they were multi-bit errors, then maybe it is a UEFI BIOS bug (incorrect memory map).

If you test one stick in a different slot do you still get the errors? (i.e. did you rule out a bad slot)?

[merczellevente]

Thanks for the prompt reply. I do not have the error logs. Yes, I did the bad slot ruling out.
I will do error logging and attach here in the next few days.
Thanks again

[merczellevente]

Hi,
So here I am.
My coding is the following:
A1 = Memory stick A into Dimm slot 1
A2 = Memory stick A into Dimm slot 2
B2 = Memory stick B into Dimm slot 2
A1B2 = Memory stick A into Dimm slot 1 and Memory stick B into Dimm slot 2

and so on. I made 10 different tests (A1, A2, B2, B3, C3, C4, D4, A1B2, C3D4, A1B2C3D4)

There are 3 photos for each result (I suppose this was completely unnecessary, but whatever). Sometimes error numbers differ because I did not set the Pass number for 2, but for 4, and I ended the test during the 3rd Pass.

Apparently, this time the address errors are the same (A10DAFFC), no matter which memory stick goes into which Dimm slot. It appears strange to me, even though I am not an IT guy.

What do you think? How is it possible that the old version of Memtest brough no error, but the new version did.

Are these 3 photos enough, or should I upload the remaining (I have 30 photos)? Also, I would happily send over the MemTest86 log, but here I cannot upload today's test results because it is 440+KB in size.

Please, help me decide whether it is a HW problem.

[David (PassMark)]

Yes it is strange. Having all 32bits in error, at a single memory address, on multiple sticks isn't characteristic of a RAM failure.

How is it possible that the old version of Memtest brough no error, but the new version did.

Lots of possible reasons. Different multi-threading behaviour, change to the algorithms, different memory map from BIOS. The latter being the most likely one to make the difference. UEFI supplies a different memory map compared to BIOS. Different ranges of RAM are available for testing.

Is there any new firmware available for this motherboard?