There's Now A Patch Adding Ryzen / AMD Zen Temperature Support On Linux

[phoronix]

Phoronix: There's Now A Patch Adding Ryzen / AMD Zen Temperature Support On Linux

Linux hwmon developer Guenter Roeck has posted a patch adding support for Family 17h (Ryzen/Threadripper/Epyc) temperature monitoring support to the existing k10temp Linux kernel driver...

http://www.phoronix.com/scan.php?page=news_item&px=AMD-Zen-k10temp-Patch

[JPFSanders]

Yay!, at last! better later than never, and with improvements :-D

[smartalgorithm]

Very nice. Hope that's going to be very safe to buy a new AMD system next year

[RavFX]

smartalgorithm You can ask AMD directly by the same support channel then when you have that segfault bug to bug CPU, if you want to but a bunch of unaffected one, they will sell them to you, pre-tested and without segfault issue.

Also, the new batch is way better for memory XMP at 3200+, I got report of system with memory issue, even getting the memory at 2933 and 1.43V... After CPU was swapped, XMP 3200 at 1.35V was just fine So new batch is also better on memory!

[Adarion]

Poor Guenter Roeck, I think he really could use some hacker's help. There is so much still open, so many chips yet to support (or even obtain data sheets usable for writing drivers). The SIO/EC and general hwmon / lm_sensors /flashrom situation does not seem to have improved in the past on that area.
But great to see this support now, maybe it was not a most horribly pressing issue, but temperature sensors and the likes are there for a reason.

[MoonMoon]

Also, the new batch is way better for memory XMP at 3200+, I got report of system with memory issue, even getting the memory at 2933 and 1.43V... After CPU was swapped, XMP 3200 at 1.35V was just fine So new batch is also better on memory!

You use a sample size of 2 to determine the quality of an entire batch? Every somewhat experienced overclocker will tell you that differences like you describe are common even between CPUs from the same batch.

[rene]

Phoronix: There's Now A Patch Adding Ryzen / AMD Zen Temperature Support On Linux

Linux hwmon developer Guenter Roeck has posted a patch adding support for Family 17h (Ryzen/Threadripper/Epyc) temperature monitoring support to the existing k10temp Linux kernel driver...

http://www.phoronix.com/scan.php?pag...-k10temp-Patch

well, yeah, Michael was crying so much and the support patch is a "few"-liner ;-)

[kaseki]

Yes, this is wonderful in principle and long awaited, and appears to tap the CPU output rather than (in the case of Asus C6H) rely on divining the mystical 8665 super I/O chip. Thank you Guenter Roeck!

However, for those of us naive in kernel grokking, some help with how to use the patch would also be wonderful. For starters, information about which kernels it is compatible with. Second, information about what, if anything, needs to be done to deal with an already embedded build of his previous contribution found at https://github.com/groeck/it87. Third, a "make" file if required for compilation, and fourth, whatever other hand-holding directions the naive need.

Or is this a patch that cannot be inserted into an already compiled kernel?

[RavFX]

You use a sample size of 2 to determine the quality of an entire batch? Every somewhat experienced overclocker will tell you that differences like you describe are common even between CPUs from the same batch.

By now, five, not two. And I send mine in a day now that I got a temporary R3.
There simply better because the process got better with time. And that's probably why the segfault bug is gone as AMD did not say what's the fsck and the stepping is the same.

On my current 1705PGT one, XMP does not work, I can run the ram at speed, but 1.45V

[haagch]

So this is the real temperature read from the CPU that was previously under NDA, right?