Memtest v7.0 BETA 1 stuck at Test 12...

[bcs]

... on a Dell PowerEdge R720 with 128 GB PC3-12800R RAM after 93 hrs:


However, after ESC'ing out and selecting 'Skip the current test' it continues with Test #13 fine.

[keith]

Do you have the logs for this run?
Is it reproducible?

If so, can you e-mail the MemTest86.log under EFI\BOOT\ of the USB drive to us.

[bcs]

Unfortunately I have no log since Memtest booted from CD-RW. I'll check if I can re-create the issue by using UFD and then just start test 12.

[bcs]

I was able to reproduce it. Here's a screenshot from this morning:


Notice how it stalls at the exact same Address and Pattern as yesterday.

Edit: Also contrary to what I wrote earlier I didn't just run #12 for this second test, but instead I ran the full test.

I'll e-mail the logs. Thanks for looking into this.

[bcs]

This issue has been investigated and a patch will be released in the coming weeks. Thread can be closed.

[dsullaustin]

BCS, what do you mean issue was investigated and a patch will be released in the coming weeks? I hope and pray that you're not a Dell employee and that a free, consumer grade memory diagnostic just identified a lockup on an enterprise server. That would be laughable, but within the realm of reality.

[keith]

The Test 12 issue seems to be fixed after disabling code optimization. It may have been the case that the SSE intrinsics that were used in Test 12 confused the optimizer, causing incorrect code to be generated. Or it might be a actual hardware bug with this machine as we haven't seen this problem in any other machine. We haven't had any reports of similar issues from users. Also, the problem seems to only occur in Parallel mode; Single, Round Robin and Sequential modes all complete without any issues.

We will release a patch in the next week or so to fix this issue..

[bcs]

dsullaustin,
Your prayers have been heard

... that a free, consumer grade memory diagnostic just identified a lockup on an enterprise server. That would be laughable, but within the realm of reality.

Nah, it didn't lock up the server as mentioned in my initial post ("However, after ESC'ing out and selecting 'Skip the current test' it continues with Test #13 fine."). It was just the test that stalled/failed/hung. But yes, I agree, the more complex these servers get, the more likely to encounter screwups.