New Build, Getting Freezing and Crashes. Advice?

Discussion in 'General Hardware' started by cybereality, Jan 8, 2018.

  1. cybereality

    cybereality 2[H]4U

    Messages:
    2,653
    Joined:
    Mar 22, 2008
    So I just built a new build and it's not stable. Running Ubuntu now, and I've gotten repeated freezing (total lockup, image stays but PC cannot be recovered without pressing the power switch) and some game crashing. The freezes have happened at different times, on a Firefox browser, just on the Terminal typing, or while compiling code. One game, DeadCore, crashes to desktop after 1 minute into a level (I tried like 5 times, same deal). I've also tried both Ubuntu 17.10 and 16.04, which froze minutes into getting into the desktop without even installing anything. I feel like the issue may be in the hardware, but I'm not sure what.

    This is what I've tried so far:

    - 24 hours of MemTest86+, completed 8 passes, no errors.
    - 22 hours of Prime95, blended stress test, no errors.
    - 2 and a half hours with Unigine Heaven running in loop, 1080p medium, no crashing.

    I feel pretty good that the RAM and CPU are OK. So it could be something else, I'm not sure what. Any advice on how to test the rest of the system? I have a spare video card, so I'll probably try that later but over 2 hours in Heaven seems like a decent test. Any other suggestions?

    Here are my specs:

    AMD RYZEN 7 1700 8-Core 3.0 GHz
    ASROCK AB350 Gaming-ITX/ac AM4
    GIGABYTE AORUS Radeon RX 580
    G.SKILL FORTIS 32GB (2 x 16GB) DDR4 2400
    SILVERSTONE SST-SX650-G 650W SFX 80 PLUS GOLD
    CRUCIAL MX500 2.5" 1TB SATA III
    WD BLACK 2TB 7200 RPM SATA 6Gb/s

    Thanks in advance.
     
  2. cybereality

    cybereality 2[H]4U

    Messages:
    2,653
    Joined:
    Mar 22, 2008
    Okay, I may have figured it out.

    Found a StackOverflow post saying you need to enable AMD mircocode firmware support in software options.

    I just checked this box, and now things seem working (at least for the past hour or two).

    AMD_Microcode.jpg

    Played about 1 hour of DeadCore and no freezing. Will keep an eye out, but I really hope this was it.
     
    Speedeu4ia likes this.
  3. cybereality

    cybereality 2[H]4U

    Messages:
    2,653
    Joined:
    Mar 22, 2008
    Actually, it's not fixed. It was working perfect for 2 days, but it's happened again. Any ideas?
     
  4. cybereality

    cybereality 2[H]4U

    Messages:
    2,653
    Joined:
    Mar 22, 2008
    So I think I finally found a solution.

    I did a number of hardware stability tests from here: https://blog.codinghorror.com/is-your-computer-stable/ and everything passed without error.

    Following this thread I set the cstate in grub: https://bugs.launchpad.net/ubuntu/+source/linux/+bug/1690085/comments/66

    Then I went into BIOS and disabled AMD Cool n Quiet and disabled C6.

    Was previously getting a 100% reproducible crash when loading a level in the game DeadCore. I was able to load a level and quit 10 times with no crashes. So far so good.

    Hopefully this was really it.
     
  5. cybereality

    cybereality 2[H]4U

    Messages:
    2,653
    Joined:
    Mar 22, 2008
    Hmm.. wait. Well those changes helped but I'm not totally in the clear. Looks like I may be affected by the Ryzen Linux segfault issue. Running kill-ryzen fails within seconds.

    I've now disabled opcache in BIOS and kill-ryzen has been running for 15 minutes without segfaults. Only other thing I haven't tried is bumping up core and soc voltages (which I'll try if the compile fails).

    Really not looking forward to RMAing the CPU if I'm still affected, as apparently some people have RMAd and got bunk chips as replacements. But I'll see how things look.
     
    Speedeu4ia likes this.
  6. cybereality

    cybereality 2[H]4U

    Messages:
    2,653
    Joined:
    Mar 22, 2008
    So kill-ryzen still failed after running for 4 hours, but there weren't any segfaults this time. Maybe running out of memory or some other issue. I'm going to play some games and other tests but maybe I'm okay.