System Instability

[Porkey]

Spoiler

CPU: FX-8370 (stock)
Motherboard: ASRock 970M Pro3
RAM: Corsair 32GB (4x8GB) 1600MHz DDR3 (running at 1333MHz)
GPU1: 980 Ti Classified
GPU2: 950 SC
PSU: Corsair HX1000i

 

It seems like there's always something critically wrong with my system, no matter how I run it.

Last year it was a faulty kit of RAM.

A few months ago I was dealing with a dying H100i, which I sent in for RMA.

 

Now, as of a few days ago, my computer has gotten much less stable. I keep waking up to find my computer at the desktop, having restarted on its own not long after I went to bed.

I come home from work to find it restarted on its own right around the time I walked in the door.
 

What is happening is the computer is not just "restarting" on its own; it is bluescreening. To be completely honest, I've been experiencing random BSODs since I rebuilt my current PC with a new motherboard (ASRock 970M PRO3), but I always chalked it up to just being "random" as I do things to my Windows installation that would warrant an occasional BSOD.

Over the past few months I noticed BSODs becoming more frequent than I would expect. Oddly, they were all memory related errors. It's fuzzy in my own memory, but usually it would have to do with a driver or program trying to access memory that it isn't allowed to.

Now in the past few days the BSODs have gotten really out of hand, to the point where I had my computer bluescreen four or five times in the space of 30 minutes before I had to leave for work. Then I got it up and running again, only to come home to find out that the computer had lasted about 4 hours before bluescreening again.

Sometimes the PC can go all night being left alone without a BSOD. Sometimes it will BSOD within 10 minutes of being restarted.

 

This sounded like another RAM-related issue, so I loaded up Prime95 and went to bed last night. Needless to say, that didn't go so well, as I came back to a blank desktop again today. Checking the log, Prime95 completed all of the tests successfully until the computer magically crashed.

Now I've tried two separate versions of Memtest86, and the test is freezing early into the test.

As I've written this post on my server VM, I've been fidgeting with it and I'm not sure if it really is a RAM problem. I also suspect both the motherboard and the CPU itself.

 

Is there anyone out there who is familiar with odd behavior of AMD's crappy hardware?

[3 Lions]

It could be a faulty RAM controller on the board. 

[Porkey]

Now that you mention it, you could be right, but I was completely certain that FX chips had their northbridge (and memory controller) on the CPU?

https://support.amd.com/en-us/kb-articles/Pages/ddr3memoryfrequencyguide.aspx#controller

[3 Lions]

2 minutes ago, Porkey said:

Now that you mention it, you could be right, but I was completely certain that FX chips had their northbridge (and memory controller) on the CPU?

https://support.amd.com/en-us/kb-articles/Pages/ddr3memoryfrequencyguide.aspx#controller

You could be right but either way that's what I'm saying, if you got a new kit of RAM last year and it's still not working properly then it could be one of two things. 

 

1) The traces on the board for the memory are damaged.

2) The RAM controller is faulty. 

[Porkey]

The kit of RAM worked well initially; I'm pretty sure I tested it in both my rig and the server (known good for memory) using memtest and a few other tools. Neither my server or my main rig can run the ram at 1600MHz, but everything from my old OC settings of ~1400MHz and below turned up no errors.

 

I don't understand how the motherboard traces or memory controller could become damaged so suddenly like this.
The motherboard is likely out of warranty, but should I try talking to ASRock anyway, seeing as this board always seemed iffy? Or should I suspect the CPU and talk with AMD?

[Porkey]

Things have just gotten weirder.

I've been experimenting with memtest86 and memtest86+ on both my machine and my server.

I've been testing two sticks at a time in my server with the version of memtest86 that comes with unRAID (5.01 I believe).

I have a dedicated memtest USB drive with version 4 and a version 7 (I think) that runs in UEFI mode.

 

I've experimented with the core options; setting memtest to use all cores, round robin, sequential, and single-core modes.

Setting memtest to use all cores on the server eventually causes it to freeze. Sequential, round robin and single core all seem to pass with no errors.

Setting memtest to use anything other than single core in either version of memtest on my machine causes it to either reboot instantly, or eventually freeze.

I am currently running the last of my tests on the first two sticks of my Dominators in the server, which have had no errors so far with sequential mode.

I am testing one of the server's 2GB generic sticks in my main machine, and it has done two full passes with no errors.

[Porkey]

It's looking like we were both right, Dario.

When my server didn't freeze while testing my Dominators, it passed with no errors.

My FX-8370 tested my server's 4x2GB sticks of ram and ended up with 4 errors after one run.

 

I then swapped my FX-8370 out for my old Phenom II x6 1045t, and it did four passes of memtest on my server's RAM with no errors.

I've now put my whole kit of Dominators back into my rig (lowering speed and timings to stay within the Phenom's even lower capabilities).

The Phenom has so far not had any errors, whereas the FX chip would instantly throw hundreds of errors and then freeze, no matter what.

 

I suspected the CPU from the start, but I think I'll be looking for a real motherboard soon either way. I'm still going to go through with all of the testing, just to be sure, as the Phenom is still only 33% into the first pass.

 

I'll probably be back with more updates later today.