I think I can finally say I have a stable system once again. After the RMA replacement of my Ryzen CPU to fix the “Performance Marginality Problem” I noticed that I was still getting some regular system lockups/freezes when it was left idle.
How long this secondary issue has been present I am unsure. Over the summer I had been using the machine but turning it off after use as it was keeping the room a little too warm. I probably didn’t notice the idle issue as I wasn’t leaving it idle. Any occasional occurrences of problems I attributed to the old CPU.
The current workaround has been to disable the C-States in the BIOS. It appears when the system is idle it is going into a low power mode but for some reason this isn’t happening correctly. At this time I’m unsure if this is related to the BIOS itself or Fedora 26 running a 4.13 kernel.
I have upgraded both BIOS (AGESA 188.8.131.52b) and kernel over the past few months so I will have to test again when one of them changes. Next week Fedora 27 is released, it will likely ship with a 4.13 kernel but 4.14 should follow shortly after. A BIOS with AGESA 184.108.40.206 might arrive before the end of the year.
For the time being I will settle for a little additional power usage and no crashing.
This week I finally decided to send away my Ryzen 1800X CPU for a warranty replacement. I had a CPU made early in the production cycle (2017 Week 12) and it was affected by the “Performance Marginality Problem”. Once it became apparent that it wasn’t going to be fixed by a bios microcode update a warranty replacement was always going to be an option.
Sending away the processor was going to be a hassle as the machine would be offline for the duration. The awkward first step was removing the heat sink to read the serial number and production date from the top of the processor. With this step done and the heat sink remounted I was able to continue using the machine until I was ready to send it away.
The steps for the RMA were straight forward. Fill in a form on the website with my contact details and the serial number of the processor. Within 24 hours I had a response, quicker than I had expected. Due to the weekend I could only arrange to post off the processor on Monday morning (Following the packaging information instructions supplied). Being in the UK my processor was sent to the Netherlands, DHL shipping and handling paid for by AMD.
Two days later I was emailed to say my processor had been received and checked. The RMA was approved. It took another day to receive a notification that a replacement was on its way and I received it a day later. In all it took about a week, the replacement I received is from 2017 Week 37. It appears the problem is solved. I am impressed with how smoothly the whole RMA process worked.