v10.2 Build 1006

[Hotwired]

We are experiencing some strange issues with the newest build (1006) that we did not experience with build 1004.

We have run 1004 on dozens of systems that are identical hardware-wise and never had a problem. This is the first time using build 1006.

We are getting RAM and GPU RAM errors that we are starting to believe are not being caused by faulty hardware. We are only testing for 2D video, CPU, and RAM. It seems after starting the test, we get memory pre leak messages in event viewer noting both the memory test and GPU test. Shortly after that, we start to get errors in BIT.

We have gone through numerous steps, including replacing hardware, reloading OS (Win 10 Pro), and running Memtest86. Memtest comes up clean after 12+ hrs / 4 passes and hardware was replaced with new and we had the same result. I understand that the new hardware that was used as replacements could be faulty as well, but if we run the same tests on build 1004, the system passes the tests.

Has anyone else run into anything like this?

[David (PassMark)]

What are the hardware specs for the machine?

Can you post the actual error messages.

Can you turn on logging in preference window and send us the log file.

[Hotwired]

ASUS Z690 MB
12900k
Corsair 128GB DDR4
2 x Samsung 980 Pro
ASUS RTX 3060
EVGA 850w PSU

Directly after I received the erorrs, I stopped the test and ran v1004 (no hardware changes) and it passed a 12 hour test.

The log file is of the passed test. I deleted the log file of the failed test, but attached an image of the errors.

Both v1004 test that passed and v1006 test that failed were run from USB flash drive (different drives). Both drives have been used to run BIT many times.

I am running BIT in debugging mode now, with BIT installed on the system (Not USB flash drive).

[David (PassMark)]

There was no changes to the RAM test between the 1004 and 1006 releases.
There was a change to the 2D test, but only to detect new nVidia TCC hardware. The actual test didn't change.

From the screen shot, they do look like real hardware errors.

Interesting experiment would be to swap back to 1006.

The 2D video RAM errors reported however might be caused by a failure of main RAM as the data is copied between main RAM and video RAM for the test.

If you swap the main RAM and test with 1006 does the error stay the same (e.g same byte / buts in errror)?

[Hotwired]

I tested in debugging mode (1006) last night, same tests (12 hours, RAM, CPU and 2d Video). It passed again with no issues.

I have made no hardware changes. I can send logs, but I don't know if they help. The only difference here is that I installed BIT locally instead of running from a USB flash drive.

Now the system has passed both 1004 and 1006 (12 hours) with no hardware or software changes since the failed test.

Is 12 hours long enough to test the 128GB of RAM?

I supposed I could have a bad USB flash drive or something going on with the USB port/controller, but it passed 12 hours of 1004 using a usb drive to run the test (it was a different USB drive).

It really is odd. I haven't seen errors in the last two test, both 1004 (USB) or 1006 (locally installed). But I have changed nothing since the failed test.

I wonder if I somehow have a bad BIT install on the flash drive that I have 1006 installed? It failed the test when running from the USB drive, but passed when it was installed locally in debug mode.

[Hotwired]

I have swapped the main RAM between the first and second failed test. It was the first thing I repalced and re-tested. I can't recall if the errors were the same. I have reinstalled the OS since the first failed test and didn't save the log.

[Hotwired]

Could this have anything to do with an unsigned driver? I do recall seeing something about an unsigned driver the first time I ran BIT (1006). I haven't seen it in the last few runs.

[David (PassMark)]

Could this have anything to do with an unsigned driver

Not really. Signing doesn't prove the driver works and it bug free.
Signing proves who (which company) the driver came from. Nothing else.

But code signing is now mandatory for drivers in most circumstances. Windows won't load them otherwise.
(But there are also different levels of code signing to make things more complex).

[Hotwired]

I understand that signing has nothing to do with the functionality of drivers. I just thought that Windows might not have fully loaded BIT or refused to use certain parts of BIT.

I ran 12 hours of build 1006 last night from the same USB flash drive as before and it passed. I am stumped.