Since I wanted to check my Ryzen CPU for the infamous segfault error I had to create a bootable Linux USB stick. I decided to go with Fedora, because @Peanut253 mentioned that Ubuntu had some kind of SMT bug.
But when I restarted my PC and wanted to boot from the USB stick I got the following error messages: [ 0.114030] mce: [Hardware Error]: CPU 7: Machine Check: 0 Bank 5: bea0000000000108 [ 0.114034] mce: [Hardware Error]: TSC 0 ADDR 1ffffbc875c00 MISC d012000101000000 SYND 4d000000 IPID 500b000000000 [ 0.114038] mce: [Hardware Error]: PROCESSOR 2:800f11 TIME 1504809452 SOCKET 0 APIC 7 microcode 8001126 [ 0.156013] mce: [Hardware Error]: CPU 11: Machine Check: 0 Bank 5: bea0000000000108 [ 0.156015] mce: [Hardware Error]: TSC 0 ADDR 1ffffbc4b3ab4 MISC d012000101000000 SYND 4d000000 IPID 500b000000000 [ 0.156020] mce: [Hardware Error]: PROCESSOR 2:800f11 TIME 1504809452 SOCKET 0 APIC b microcode 8001126
After this the system reboots automatically; the Ubuntu Live Stick reboots without any error messages.
I don’t have any problems in Windows and I even used Handbrake to encode some videos for two hours…
Any idea what’s wrong?
My System:
Ryzen 1700X
ASRock X370 Taichi (Bios vers. 3.00) <-- upgrade to 3.10?
2x Ballistix Sport LT 16GB Single DDR4 2400 MT/s (PC4-19200) DIMM 288-Pin Memory - BLS16G4D240FSE
Corsair RM1000i
MSI Radeon r9 390oc
Which version of fedora? First thing to do is make sure you’re on a supported kernel version. (4.11 or later)
This appears to be a hardware MCE, which could be from an unstable overclock (CPU or RAM), so I’d start by backing them down if you’re running above stock speeds.
Hmmm, that should be fine. I remember Wendell’s had success with F26.
If you haven’t already, upgrade to the latest firmware available.
Good.
I’m assuming you’re able to reproduce the issue. Sometimes this can happen and just be a fluke.
I’ve just found that your RAM is not on AsRock’s Memory QVL. While I’ll continue to help you, the most likely candidate is that your memory is incompatible.
Yeah, I tried it four times with Fedora and two times with Ubuntu 17.04 (Ubuntu’s kernel could be too old for Ryzen I suppose)
This might be the issue. I picked up the RAM when it was on discount before Ryzen’s release. I figured any RAM should be fine and I didn’t know that Ryzen runs faster with better RAM. Unfortunately, there is no easy way to test this, since I don’t know someone with compatible QVL RAM. (Might this be covered by the warranty? Amazon is pretty lenient…)
If you got this from Amazon, I’d try to send it back saying it’s not working and make sure you pick something up on the QVL. I’d aim for something like this:
It’s going to be a bit more expensive, but if you can afford it, you’ll be able to bump your memory a lot higher and it’s on the QVL.
Thanks for taking the time. I’ll still install the new BIOS first and after that I’ll decide my next move. It also depends on how Amazon reacts, but as mentioned earlier they seem pretty lenient. One problem that remains though is that I’ll habe a few days without any RAM and therefore without my desktop PC. So, I’ll need to time that properly
try noapic on the kernel command line. use this guide substitute noapic for acpi=off.
this is a one time deal, to permanently make the change (if it works) you will need to edit /etc/default/grub then run update-grub.
For Ubuntu, you can always disable SMT, update the kernel to the latest version and then reenable SMT. A cold-boot is required to reenable it. Doing that is annoying, which is why I recommended Fedora since that one worked out of the box for me.
But if Fedora doesn’t work, because Linux, then try some other distro like Ubuntu/Suse or w/e. It just needs to be stable enough for you to crash it later with the ryzen-kill script. Technically, a live-usb should also work so you don’t need to actually install anything.
Edit: Those are MCE errors. MCE errors were the second type of error people were reporting on defective Ryzen CPUs. Totally missed that due to the lowercase “mce” but yes, that alone makes the CPU worthy of an RMA, after doing the basic troubleshooting steps: UEFI update, reload uefi defaults, 2133Mhz Ram etc.
This. Although the issue was overshadowed by the segfault issue, many users on AMD’s forums have reported these MCE errors, and I would highly suggest RMAing, after you verify that your other hardware isn’t the cause (UEFI update, JEDEC 2133MHz RAM, no overclock).
After further testing I decided to contact Amazon in order to RMA my CPU. I told them about the MCE error(s) I was getting and Amazon quickly sent a new one, so I’m supposed to return the old one now. However, when inspecting the serial number of the new CPU I discovered it to be on the other’s post list in the “Chips With Issues” section:
Should I test the new one (and as a consquence possibly test a bunch of others), or contact Amazon again?
Personally I am not accepting a CPU before 1725, simply because I don’t want to have to deal with this stuff again. I already sent back a 1600X from week 15 or so. Good thing you can look at that number without breaking a seal.
So in this case, a RMA for my CPU should be in order (and not my RAM).