Narrowing down ECC testing to the area where problems found isn't working correctly..

[SnakeByte]

Both Memtestx86 6.3.0 Pro and the 7.0 Beta 1 find "ECC Errors Detected" using Test 9 at Addr: 0x236B0BA58 on one of my servers. It's always only Test 9 that finds the problem, and always at that specific address.

Since I started this test after the server unexpectedly rebooted, I'm assuming that the ECC is sometimes failing to correct the error, so I wanted to adjust the testing to focus on that one point of potential failure.

I thought if I changed the tests to only use Test 9, and narrowed the memory range to 0x236B0B000 - 0x236B0C000, and cranking the Pass number up to 999, I'd see LOTS of matches. However, I'm up to Pass 43 now, with nothing found. What's going on?

[keith]

Can you upload or post a snippet of the MemTest86.log file located in the EFI\BOOT\ folder of the USB drive.

Memory addresses aren't completely independent of each other which means that errors can also show up by writing to other memory addresses as well. So you may need to increase the range in order to catch the error.

But if you are getting ECC errors, it may be a sign of faulty RAM. So it may be a good idea to find replacement RAM and see if the symptoms still occur.

[SnakeByte]

Can you upload or post a snippet of the MemTest86.log file located in the EFI\BOOT\ folder of the USB drive.

Memory addresses aren't completely independent of each other which means that errors can also show up by writing to other memory addresses as well. So you may need to increase the range in order to catch the error.

But if you are getting ECC errors, it may be a sign of faulty RAM. So it may be a good idea to find replacement RAM and see if the symptoms still occur.

I've been burning the iso to a cdrom. Is the log still retrievable?

[keith]

The log is only available when running from a USB drive. We recommend using a USB drive over CDROM because some features require the boot media to be writable to function correctly