Previous installments:
AMD Ryzen
problems, Part 1
AMD Ryzen
problems, Part 2
For the same system in a previous article, AMD has been very helpful, and they have talked me through working on making sure that I have optimal cooling and that my AGESA version is sufficiently new.
The system had been exhibiting a "Linux Marginality" problem described by Michael Larabel of Phoronix. My Debian GNU/Linux system would sometimes generate errors under highly parallel loads, such as compiling software; this problem seems to affect early Ryzen chips manufactured before week 24 or 25 of 2017. Mine had been manufactured in week 5.
The conclusion to the problem? AMD Issued me an RMA. They also paid shipping both ways. This doesn't seem to be standard practice, and I appreciate the tech that I worked with going to the extra trouble for me. AMD received the RMA processor at their center in Miami, FL on March 30th, and I received the replacement on April 11th, 2018, which I thought was pretty good turnaround time.
The new chip, stamped with date code 1737, meaning week 37
of 2017 (September), is a gem and a wonder. The previous chip
would tend to run near its thermal limits (around 68C or 69C)
*
under full load, whereas the replacement chip will barely
reach 64C. Plus, of course, no evidence of the previous problems.
PROBLEM SOLVED with the replacement processor. And lesson
learned
reinforced about buying new technology at its first
release (I had ordered the Ryzen chip within a few days of its
launch day).
I posted a few videos of the test on the old chip in Part 1 and Part 2 of this article, but it doesn't make sense to post a video of the new chip *not* failing; it would simply be a very, very long video of nothing happening except a Ryzen properly compiling the GCC suite.
I want to thank Mr. Larabel for helping me to understand what was going on through the high quality reporting on his website, and especially AMD for being friendly, helpful, and effective.
As borne out by my media encoding benchmarks and by synthetic benchmarks, the Ryzen 7 1700X is fully 100% faster than the AMD FX 8300 that I bought it to replace, so even with the brief timeout to RMA and with the learning experiences, I consider it a good deal overall. The Ryzen even keeps up with a 10 core Xeon I also have. It's not only cost-effective, but *fast*.
* The previous videos show much cooler temperatures because the chip, a 3.4 GHz Ryzen 7 1700X, was clocked through BIOS at only 3.0GHz, to eliminate questions over whether the failure was due to heat, or due to some other internal cause.