Thousands of errors - false positives?

[nexigen]

Using memtest86 v6 free edition and booted it off my thumbdrive. This is new ram, works fine in games and on windows10, but fails with memtest. I get -thousands- of errors at step3 (inversions, ones and zeros) and shortly after step4 begins it reboots. I get this even at the default speeds, volts, and timing.

Side-note I set the core usage to all 8, parallel. Any ideas what's going on with this, or if I should use something other than memtest?

I get the same results regardless of timings and overclocking, it defaults to 1333 and I've also tried 1600 and 1866mhz with adjusted timings. The default 1333 also gives horrible timings, 11-11-11 compared to the advertised 1600 9-9-9. All of them were running the default 1.5volts and all gave tons of errors at that same part of memtest.

The memory is GSkill Ripjaws 1600mhz - F3-1600C9D-16GXM
Using Asus M5A97 Plus board with amd fx-8320 8core.
Board is running the latest bios. I also tried using BurnInTest from passmark - and tested only the ram at the default 50% load. It passed 2 cycles without any errors.

Test Duration:
The thousands of errors were on the first cycle, part 3, about 30-35 seconds in. I think it finished step3 (just a few seconds) and had 350,000+ errors.


The Log File:
I just went through the log and here's a couple lines that might clear up what's happening and maybe you can tell me how to fix it?

2015-12-03 21:57:02 - RunMemoryRangeTest - CPU #2 timed out, test time = 0ms (BSP test time = 708ms)
- The above one repeats a lot, not just cpu#2 but all of them. Shows up about 20 or so times in this test. Wondering if the cpu's are overloading the memory and timing out is causing the errors?

2015-12-03 21:57:05 - MtSupportRunAllTests - Test execution time: 25.666 (Test 3 cumulative error count: 16246213)
2015-12-03 21:57:05 - GetAMD15Temp - Temperature: 27

Test#4 (Moving inversions, 8-bit pattern)
It repeated the following a few times (different cpu/apu #'s) before it rebooted the system/crashed.

2015-12-03 21:57:06 - RunMemoryRangeTest - CPU #7 completed but did not signal (test time = 27ms, event wait time = 1001ms, result = Success) (BSP test time = 28ms)
2015-12-03 21:57:06 - RunMemoryRangeTest - Could not start AP#7 0x0000000019100000 - 0x000000001D100000 (Not Ready). Resetting...

[keith]

Can you upload or e-mail us the entire log file. Our contact details are here.

I suspect it could be a UEFI BIOS bug with the multiprocessor service implementation. So I would try running in single CPU mode to see if the errors still appear.

[nexigen]

Can you upload or e-mail us the entire log file. Our contact details are here.

I suspect it could be a UEFI BIOS bug with the multiprocessor service implementation. So I would try running in single CPU mode to see if the errors still appear.

I can send it if you'd like, but I'm already ahead of you. I just finished a full cycle with 0 errors on one cpu. I feel like the multiprocessors overloaded the ram, causing the datastream to timeout, resulting in errors. My guess is it took 3-4 tests before the ram fully overloaded, I guess I could try to test that theory by skipping the first 2 tests and setting the cpu's to parallel to see if it's still that same test that fails or if it would be a couple downstream, which I may do later if you'd like.

So - if this is a bug is there anyway to avoid it and still use all the cpu's? I stopped the test at 54 minutes and it was 75% on the second cycle. That's like forever.

[David (PassMark)]

There really isn't any such thing as overloaded RAM. You pick the speed and the timings and it either works or it doesn't. The CPU can't push the RAM beyond the speed that has been set.

My likely it is what Keith suggested, it is a bug in the UEFI BIOS that causes mis-behavior with multiple threads.

So if you could send in the log that would be good.