Update: received a second replacement unit which is a 1734SUS. This one hasn't segfaulted in 1 hour of kill-ryzen (stopped because no more space left in /tmp) and a rebuild of gentoo @world set, so looks to be bug-free. At last!
Went through the RMA process started in January after sending this photo to support. They confirmed it's one of the seg fault CPUs for running Linux. AMD confirmed their receipt of the CPU and it had passed on Jan 29. But, still no replacement has come. Limping along with an old machine and a laptop. Can't believe it's taking this long.
I'm keen AMD fan, but sadly I ran into Ryzen issues.
I'm using FreeBSD. The issue occurred during source based installation, more specific - compilation of lang/gcc6 (and lang/gcc5) port.
The system just hangs or sometimes segfaults.
I've tested with different RAMs, Mainbords and PSUs and I can confirm that if I put the harddisk in another system (with older AMD CPU) the build completes successfully.
My first processor was AMD Ryzen 3 1300X (YD130XBBM4KAE), UA 1722PGT
It was replaced by AMD with the same unit with code UA 1741SUS. The result was exactly the same.
I found and tested with YD1200BBM4KAE / UA1737PGS with same failure.
Now I'm totally disappointed and I have no spare time for useless tests... Note that I haven't used kill-ryzen or other strange tests. The issue occurs during normal operation.
The BIOS is updated, with default settings (to avoid any overclocking issues). The OS is also updated and tests were made with and without latest microcode updates.
I'm not sure should I fill another RMA request? Any advice?
"Note that I haven't used kill-ryzen or other strange tests."
kill-ryzen is not a strange test. My advice is to try that script on the 1741SUS processor. The other 2 processors likely have the error, but 1741SUS is quite new and should be problem free. Make sure you have enough RAM when running that test inside the RAM, else run it on the hard drive.
Also run a memory stress test, to make sure there's no memory instability problem.
Your error can have many reasons, including an incompatibility between your OS and your hardware.
I'm just starting the RMA processes for a Ryzen 5 1600 with a date code UA 1748PGS that I bought early last week.
Running "kill-ryzen.sh 4 4" it would fail in < 200s on RAM disk and "kill-ryzen.sh" in < 45mins on SSD.
I've had to order another AM4 cpu just to tide me over.
I have just purchased a Ryzen 2700 and 16GB of Unbuffered ECC. I plan to test soon..
kill-ryzen.sh fails to build under ubuntu-18.4 (with a compiler error on a xeon). I will test the pull request.