Yes, I received an older AMD Ryzen 7 1700X CPU from Amazon. It’s RMA time!
In Episode 27, I wrote about intermittently getting segmentation faults on my new AMD-based PC whenever I did Yocto Linux builds. I found some information online in New Ryzen Is Running Solid Under Linux, No Compiler Segmentation Fault Issue and AMD Replaces Ryzen CPUs for Users Affected By Rare Linux Bug.
It seems that earlier production runs of the chip had a problem with cache coherency. So, after I returned from my vacation, I put in for an RMA number from AMD, and proceeded to remove and box up the faulty CPU. Whenever I’ve done PC builds, installing the CPU and heatsink are always the most hair-raising part, and taking this apart was a little nerve-wracking. Nonetheless, after CPU was removed and the thermal paste wiped off, the markings did in fact tell the tale:
The “UA 1709PGT” on the second line designated that this was a work week 09 production run, and all CPUs prior to WW25 are expected to have the segfault issue.
So, it’s back to AMD for the bad CPU. So far, their RMA process has been pretty responsive. I expect the new CPU back next week, and will keep you posted.
Once I get my machine back together again, I’ll resume my efforts on the MinnowBoard Turbot, to install a working implementation of Linux. I expect to do some basic source-level Linux kernel debug using SourcePoint. Also, I’m planning on further investigations into the Portwell Neptune Alpha board, which supports the ASPEED AST2500 BMC, our development platform for the ScanWorks Embedded Diagnostics JTAG-based run-control debug product.