skyline7349 Posted November 1, 2019 Share Posted November 1, 2019 So a while back I build my server with Ryzen processor and to get it to work they recommended turning C State off. Is this something i should turn back on? Someone brought it up to me the other day. Quote Link to comment
John_M Posted November 1, 2019 Share Posted November 1, 2019 Which processor are you using? Only the 1000-series is affected and it was fixed a long time ago with a new BIOS option for the Power Supply Idle Control. So see here: Change it to Typical Current (not the default, Low Current) and it will prevent the cores from dropping so deeply into sleep that they can't wake up again, and re-enable C-states. For 2000- and 3000-series chips it shouldn't be needed but it won't do any harm. Quote Link to comment
JorgeB Posted November 2, 2019 Share Posted November 2, 2019 10 hours ago, John_M said: Only the 1000-series is affected Are we sure about this, I remember seeing various 2nd gen models still being affected, and I believe even a 3rd gen. Quote Link to comment
John_M Posted November 2, 2019 Share Posted November 2, 2019 3 hours ago, johnnie.black said: Are we sure about this, I remember seeing various 2nd gen models still being affected, and I believe even a 3rd gen. Well, it has never been officially acknowledged but I believe it was fixed quietly in Raven Ridge (launched February 2018), with the Power Supply Idle Control setting becoming the standard fix for earlier Summit Ridge chips. The discussion surrounding the subject is extremely muddled, with people regularly going off at tangents, as in this example: https://forums.anandtech.com/threads/ryzen-random-crashes-in-windows-10-at-idle-seems-to-be-very-common-out-there.2568509/ from which the conclusion might be that not only is Pinnacle Ridge also affected but Intel's Bay Trail too. I must have read over a hundred threads like that one since the issue first became known and none of them are definitive, including the "mega-thread" that Tom refers to. My own small scale testing (less than 100 samples) isn't definitive either, but I've only been able to observe the phenomenon in CPUs with a 2017 date stamp (presumably 2016 too but I've never seen one that early). I have a late manufactured 1800X (date stamp 2018 week 6) that is immune and an early 1700 (2017 week 4) that isn't. The 1700 is fixed by the Power Supply Idle Control setting. So my advice it to point people to where to find it in the BIOS and tell them that it is very likely to help with a 1000-series, is unlikely to be needed for anything else, but won't hurt anyway, and not to disable Global C-states. If anyone feels that advice should be modified then I'm always willing to learn from new evidence. Quote Link to comment
Recommended Posts
Join the conversation
You can post now and register later. If you have an account, sign in now to post with your account.
Note: Your post will require moderator approval before it will be visible.