Ryzen C-State related problems -- what is the root cause?

[RipClaw]

I would like to know why is that we are advised to disable C-State feature in the BIOS, to improve stability and such. For instance I do find Unraid, working good after disabling C-state.
So what is going on ???
a)
Is the issue with Ryzen silicon OR
is it with the loadable microcode OR
Is it with kernel of the operating systems, not yet adapted to utilized the C-states feature properly ??
b)
why would BIOSs ship with them enabled by default, if they know things may not work properly for many ???!!!

[Mastic_Warrior]

ACPI implementation is horrible because vendors use the spec as a suggestion rather than a guideline.
Power delivery issues possibly.
a.3 possibly due to the newness of the platform and the fact that AMD has spread themselves so thin with all of their initiatives that Intel optimizations seem to perform better than AMD optimizations (gcc, llvm, arch, etc).

[RipClaw]

Let me hope that it can be fixed in software …
Wish we had some kind of processor test/diagnostic board like the one below, to test them all!
Has anybody found AMD Ryzen’s errata document ??

[catsay]

I wish. Currently the BKDG (Bios and Kernel Developers Guide) is still only available under NDA.
The public stuff is comprised of the 17h documents on here:

https://developer.amd.com/resources/developer-guides-manuals/
Which Includes the Registers reference, the Processor Programming Reference and Software optimization guide. Some prior documents still apply such as for the PSP.

Some reverse engineered stuff on the Ryzen Pipeline here:

http://agner.org/optimize/microarchitecture.pdf

And whatever I’ve previously dumped here:

Secure Virtual Machines and Memory Encryption on Desktop Ryzen?

Looks like someone actually did the homework and read the linked papers Links from the previous thread: https://lwn.net/Articles/685215/ AMD Memory Encryption whitepaper: (Currently down) http://amd-dev.wpengine.netdna-cdn.com/wordpress/media/2013/12/AMD_Memory_Encryption_Whitepaper_v7-Public.pdf AMD64 Architecture Programmer’s Manual: http://support.amd.com/TechDocs/24593.pdf SME is section 7.10 SEV is section 15.34 Support for the extension exists within the kernel https://patchwork.kernel.org/patch/9612007/ AMD Memory Encryption Tutorial ISCA 2016 https://sites.google.com/site/metisca2016/ The Intel Equivalent technology https://software.intel.com/en-us/blogs/2016/02/26/memory-encryption-an-intel-sgx-underpinning-technology Regarding SEV A Whole dump of everything else Ryzen technical

And probably a good deal more stuff I’m forgetting about.

But no errata spec/BKDG.

Most of the errata (atleast the fixed ones) are currently best found by looking at the linux kernel.

[RipClaw]

Errata for Ryzen/Threadripper/EPYC has finally come now…

55449_1.12.pdf

There are a lot of “Suggested Workarounds” presented!
Note sure if System Software meant the Linux kernel or AGESA or either of them, depending upon the issue.

[catsay]

Hallelujah!!

@noenken - This means you may soon get some stability improvements for your Raven Ridge Laptop.

[noenken]

It is actually running perfectly fine for a few weeks already.
And yes, C-States again. Motherfucking C-States.

Ryzen/Vega laptop PCIe Bus Error

OK, pretty much a full day uptime, 12 hours videoplayback, three of those Ryan’s PGP stream, couple hours mprime -t, couple hours unigine-valley and now idling most of the time. I wouldn’t call this completely fixed because it still needs the flags to work and it also drains more power than it would without them. But for now I would say this is the solution for anyone on mobile RR. Add "rcu_nocbs=0-7" and "processor.max_cstate=1" in grub.

[RipClaw]

How much watts would be the power delta that we are talking about ?