Memory errors only after rebooting from Windows. Not on cold boot.

[David (PassMark)]

We had a customer report memory errors in MemTest86, but only after first booting windows.
If testing was done directly from a cold boot, there was no errors.

System specs
Intel Core i9-10980XE overclocked to 4.5 GHz, Gigabyte X299X DESIGNARE 10G motherboard, 4x32 GB Samsung M378A4G43AB2-CVF DDR4

Manufacturer	BOXX Technologies
Product Name	APEXX X3_03
Version	-CF

Manufacturer	Gigabyte Technology Co., Ltd.
Product Name	X299X DESIGNARE 10G

Error details

Lowest Error Address	0xC49883DF0 (50328MB)
Highest Error Address	0xC49883DF0 (50328MB)
Bits in Error Mask	00000000000003FF
Bits in Error	10
Max Contiguous Errors	1

Errors were mostly in tests 11 and 12 and happened in 10bits at a single memory address.


Speculation as to the cause

Speculation is that maybe booting into Windows enabled some piece of hardware (or changed the setting of some hardware). This hardware then started updating this single memory address, vias memory mapped I/O. Maybe a bit coded status field of some sort. So once enabled it stays enabled, corrupting RAM, unless there is a cold boot. Which implies this is really a BIOS bug as the device should have reserved that RAM from the start by use by this device.

If anyone has a similar problem let us know.

[timjohans]

Hello,

I have also purchased this board and can confirm the above. I have the following:

10980XE overclocked to 4.7GHz, Gigabyte X299X Designare 10G motherboard running latest F3c bios.
Corsair CMW32GX4M2E3200C16 memory at 3200MHz.

When cold booting (power off to PSU), running both Windows Memory diagnostics and Memtest returns no errors. However, when performing a warm reboot and then retesting, I get no errors in Memtest, however get errors in Windows Memory diagnostics.

I also get errors when trying to run XMP in both scenarios (again cold reboot is fine) - currently I have entered these settings in manually with XMP disabled.

No BSOD experienced during operations. Prior to this board, I previously had an earlier MSI X299 MB from ~2017, and upgraded to this to benefit from the better VRM's on it amongst others.

Also, when running UserBenchmark, I get an error when testing WARN: Unable to process RAM module 0h - this is running in Windows. This was not experienced on my previous MB as I have moved the compoents over to the new one.

I gather that this is a BIOS related issue? Are there any implications for day to day use? Is an update on the horizon do you think - will report this to Gigabyte.

Also, as a sidenote, I have found that this board also has a confirmed memory training voltage issue as per the below:

https://www.youtube.com/watch?v=rYuTY1MWyBM

Any further information would be much appreciated.

Best Regards,


Tim.

[David (PassMark)]

Thanks for the additional information. We don't know if Gigabyte are aware of the problem or if they are working on a solution.

[timjohans]

Thanks - I have contacted them, but they seem fairly unconcerned about it to say the least. I am wondering if it might be possible if your user with the issue could also contact them?

Is this essentially something to worry about during normal operation I guess?

[David (PassMark)]

I think it is potentially serious. But I am not surprised that Gigabyte don't care.

Added comment: Is related to this similar post
https://www.passmark.com/forum/memte...-memory-errors