V5.1.0 test fail on Bay Trail

[stevensxl]

Hi，
My Platform is Bay Trail J1900
Hynix 2GB 1Rx8 per channel
Tested v4.3.7 on 32bit BIOS, v5.1.0 on 64bit BIOS
Test 1: Two channels populated the same Hynix memory, with 64bit BIOS and run Memtest86 v5.1.0. Choose single CPU or Parallel in CPU selection. Test will fail at the step test 2, find hundreds errors. Log file as below.
https://drive.google.com/file/d/0Bxm...it?usp=sharing

Test 2: Two channels populated the same Hynix memory, enable legacy USB in 32bit BIOS and run Memtest86 v4.3.7. Can complete all test as choose single CPU or Parallel in CPU selection, no error found.

Test3: Single channel populated with the Hynix memory. Test complete under 64bit and 32bit BIOS, no error.

Do you have any idea why would Test 1 fail? Is it a hardware issue?

[keith]

Thanks for the details.

According to the log file, it seems that your system was running the 32-bit version of MemTest86 v5.1.0 which implies that it was running 32-bit UEFI. The maximum physical address is 0x140000000, which is greater than the 32-bit address space supported in the 32-bit version of MemTest86 v5.1.0, possibly causing the errors you are experiencing.

What you can try is setting the memory address limits to 0x0-0x100000000 in the MemTest86 main menu and see if the problems still occur.

Is there a reason why you are running 32-bit UEFI instead of 64-bit UEFI?

Also, what is the CPU on your system?

[stevensxl]

Hi Keith,
Thanks for your help. The CPU I am using is Intel Bay Trail J1900.
Your workaround works, and now I changed to 64bit BIOS, and got a new issue. The test will hang after pass 18% of "Test 2". Please get my screen shots and log file from following link, details are inside.
https://drive.google.com/file/d/0Bxm...it?usp=sharing

[stevensxl]

Hi Keith,
The tests we are doing, the first choice is to set Parallel in CPU selection. So the hang on Test 2 issue only happens when Parallel is selected. I also found this issue only happens with quad core CPUs, like J1900 and E3845. I also tested E3815 and E3825, both passed the test. So is there a limitation for the tool to work with quad core CPUs? Or any workaround I should do? Thanks.

[keith]

There seems to be an access restriction on the link you posted, so we can't download your log file (it isn't public).

The freeze in parallel mode may be due to BIOS firmware issues, but will need to have a look at the log file to be sure.

[stevensxl]

Hi Keith,
I reset the link, it should be public now.
https://drive.google.com/file/d/0Bxm...it?usp=sharing

[keith]

Thanks, we were able to access the log file. Can you give this debug build a try and run the same test in parallel mode as before:

http://www.passmark.com/ftp/memtest8....1.1b.1005.zip

This build will confirm whether or not the other processors were able to run the tests parallel.

[stevensxl]

Hi Keith,
Quad Core CPUs can not run this tool, will hang when I press start after selected parallel mode, but can run in single core mode. Dual Core CPUs have no such issue.

[David (PassMark)]

Can you post the log file. Thanks.

[stevensxl]

The latest log is here:https://drive.google.com/file/d/0Bxm...it?usp=sharing

[keith]

Thanks, however the log only contains details from start-up but nothing about actually running the tests. If possible, can you re-run test 2 in parallel mode with the USB drive plugged in so that the logs can be written into the file.

[stevensxl]

Hi Keith,
There's no running information about the test in the log file is because when I use the debug version provided by you, can not start the test. After I set CPU selection to Parallel, and press "S", Memtest86 hang, cannot start the test. But if I choose single in CPU selection, test still can run, this issue also will not duplicate on Dual Core CPUs.

[keith]

Thanks for the info. Please give this build a try at your convenience and send the logs as before:

http://www.passmark.com/ftp/memtest8....1.1b.1009.zip

There are no fixes in this build, just additional debug information being logged.

[stevensxl]

logfile: https://drive.google.com/file/d/0Bxm_1t0QPEJIUWJ2cGRzMUNRRW8/edit?usp=sharing
screen shot: https://drive.google.com/file/d/0Bxm_1t0QPEJIQ2NiQlVmQ1EtZUk/edit?usp=sharing

[keith]

Thanks for the files. It is likely that the reason for the hang is that the UEFI firmware did not properly signal that the processors have finished execution, causing MemTest86 to wait indefinitely. We have implemented a possible workaround. Please give try this build a try and post the logs as before:

http://www.passmark.com/ftp/memtest8....1.1b.1010.zip