1078 Errors so far with Memtest86. WHAT DO I DO?!

[FireEagle244]

I have a system (ASUS - ROG STRIX B550-F GAMING, 2700x CPU, Samsung 970 Evo 500GB, 32GB RAM, and Running Windows 10)

I've been using it as a server computer to host Minecraft servers and other things for friends. While using it, I have found that sometimes it will just..."reboot" without shutting down and restarting as you'd typically expect. It just seems to close everything and go to the lock screen. I verified Windows and am currently running a Memory test using Memtest86.

There are 1078 Errors so far (91% through the 3rd pass out of 4).

2x 16GB DDR4 3200 MHz Sticks from Silicon Power. They were cheap but I can't imagine that this is normal.

What should I do?

 

[Levent]

Disable XMP and test again.

[FireEagle244]

3 minutes ago, Levent said:

Disable XMP and test again.

I don't believe that I have it enabled but I will check after the test is complete. 

[PDifolco]

Replace your RAM, there's no fix for bad RAM

[RussianSalad]

Adding to the recommendation of disabling XMP, sometimes reseating the memory modules and cleaning the contacts in the motherboard and memory modules themselves helps with this kind of errors. And on a first gen Ryzen machine I used to have reseating the CPU solved all the random freezing issues it had. Although I suspect that was because the computer got quite banged up when it was checked out as luggage on a plane trip .

 

Edit: from looking at the picture it appears that the same memory addresses fail over and over again. If you don't want to replace memory there are ways of telling memtest86 to parse those faulty addresses as a kernel boot argument to tell windows/linux to ignore those addresses. It's not as good as straight up replacing the memory modules but for a non critical machine it's worth a try. It's a free "repair" after all.

[FireEagle244]

8 minutes ago, RussianSalad said:

Adding to the recommendation of disabling XMP, sometimes reseating the memory modules and cleaning the contacts in the motherboard and memory modules themselves helps with this kind of errors. And on a first gen Ryzen machine I used to have reseating the CPU solved all the random freezing issues it had. Although I suspect that was because the computer got quite banged up when it was checked out as luggage on a plane trip .

 

Edit: from looking at the picture it appears that the same memory addresses fail over and over again. If you don't want to replace memory there are ways of telling memtest86 to parse those faulty addresses as a kernel boot argument to tell windows/linux to ignore those addresses. It's not as good as straight up replacing the memory modules but for a non critical machine it's worth a try. It's a free "repair" after all.

I'm definitely going to try all of the free options before getting new ram as I don't have much money that I can spend at the moment.

[FireEagle244]

10 minutes ago, PDifolco said:

Replace your RAM, there's no fix for bad RAM

Seems to be really bad, from my perspective with my limited experience testing ram. 1078 errors sound significant to me. 
I'm going to test some things like @RussianSalad and @Levent suggested first.

[RONOTHAN##]

Try lowering the memory speed to 2933 and trying again. You're on a 2700X, which those had pretty terrible memory controllers. Most could do 3200 like you're trying to run, but not all of them could and it's possible that even JEDEC 3200 is too high a memory speed for your CPU. 

[FireEagle244]

5 hours ago, RONOTHAN## said:

Try lowering the memory speed to 2933 and trying again. You're on a 2700X, which those had pretty terrible memory controllers. Most could do 3200 like you're trying to run, but not all of them could and it's possible that even JEDEC 3200 is too high a memory speed for your CPU. 

Im running a memtest with 2933MHz currently and it is looking good. I turned off XMP/DOCP (TBH didn't know it was on since it was called DOCP) Going to run it at 2933 for a while so that I can get the game servers back up. Going to test a little higher without XMP when there are less players wanting to play