I am having a problem with my Ryzen 7 1700 upgraded FreeBSD server. I recently upgraded two servers from older 6 core AMD processors to new Ryzen 7 1700 processors. Since replacing the hardware, the servers have been extremely unstable. Uptime between crashes ranges from just a few minutes to just under 17 hours. The older hardware was rock solid. This occurs on the two systems upgraded to Ryzen, so it is not a memory problem.
Hopefully I can find some help for this problem here.
I disabled SMT on the CPU and got the same results. The crash is always random. I can't believe AMD would be so disinterested in these problems. I spent a lot of money on this hardware and it is currently a paperweight. I understand there are severe stability problems with Ryzen in BOTH linux and BSD Unix.
Ignoring the problem won't solve it.
Running the system live as a router/firewall, I now have a whopping uptime of 5 minutes. I tried another motherboard with the same results. I sent a request direct to AMD concerning this ongoing issue and have not yet received a response! I think AMD may know how bad this actually is and is ignoring the issue instead of stepping up and fixing the problem with the CPU. I am now wishing I have never started this hardware upgrade. I have wasted lots of money and time on this and AMD is silent on the matter! Very disappointed.
Not sure if it is the same as in linux during massive parallel compilation (segfaults).
MCE may be different story (it is rare compared to segfaults).
You may try disabling opcache - this solution works for most. You may try to raise SOC voltage and do not OC RAM. Or bump core voltage a bit. But all of those are workarounds only.
Anyway - contact AMD for RMA. They will replace your CPU with selected, tested chip free of problems.
I have contacted AMD for this issue (support/RMA) and in over a week, I don't have a single reply!
Processor is less than 15% load when this occurs. Since this is a routing setup, I'm sure there are plenty of parallel tasks running for both routing and firewall. About 800 clients connect to this router setup with a maximum throughput exceeding 700Mbps at peak usage.
Also, it concerns me that I have seen reports of replacement CPUs having the same issues as the returned ones.
For now I will have to keep relying on the old 6 core athlon 3200 setup and I figure I have lost lots of time and money (unless AMD can actually fix this problem).
why not check on the freeBSD forum and see if anyone there has similar problems with modern processors
might need a more recent compiler to better support the modern CPU cores better
I have done so. Plenty of reports of stability problems and a few workarounds that don't quite work. All indications point to a serious flaw with Ryzen chips.