Hi!

 

I have a Ryzen 3970X on a Gigabyte TRX40 Aorus Xtreme with 128GB Samsung ECC.

 

I have problems with random "crashes" where the system becomes totally unresponsive and i cannot find anything in the logs (I have a separate sys-log server).

 

When I bulit the system some 18months ago I did run a 48hr memtest without no errors, the system has been unstable ever since the build. The "crashes" are very intermittent, but always happen within 14 days after a system restart (can be 1 day, can be 13).

I found some post here on the forum saying some Ryzens have som bug that can cause craches, and in these cases a powerline setting can fix the problem and sometimes one have to disable the global C-states.

 

So I tried the powerline setting for a while but it did not help, so today I decided to test to disable the global C-states, but that had a very negative effect on my Windows VM's, they got really bad performance wise, and each machine takes about a 100% of a thread in the system constantly and the machines themselves are very "laggy".

 

Does anybody know if there is a way to fix this?

2 hours ago, Koenig said:

can fix the problem and sometimes one have to disable the global C-states.

You don't need to disable C-states, just set the correct power supply idle control.

1 hour ago, JorgeB said:

You don't need to disable C-states, just set the correct power supply idle control.

Sorry for not remembering the name of the setting right, but that is exactly the setting I meant when I wrote "powerline"-setting, so I have already tried that and it did not solve the problem, I changed that setting a couple of months ago and I still have had the system go unresponsive to the point where I have to make hard reset, with a maximum uptime of 14 days.

 

+ That was the exact post I was reffering to in my previous post.

 

The lockups are the root of my problem, but now that I have disabled global C-states (to try if that helps with the lockups) my Windows VM's performance i so slow it is unbearable so any help in understanding why that is and if there's anything to do about it would be much appreciated.

 

14 minutes ago, Koenig said:

so I have already tried that and it did not solve the problem

It should, make sure it's running the latest BIOS, if not it's a board problem, that setting, if implemented correctly, is enough to solve the C-states issue, avoiding the need to completely disabling them.

On 12/19/2021 at 1:53 PM, JorgeB said:

It should, make sure it's running the latest BIOS, if not it's a board problem, that setting, if implemented correctly, is enough to solve the C-states issue, avoiding the need to completely disabling them.

But it didn't, and yes when i changed the setting to what was recommended in that post you linked to, I had the latest BIOS (I updated just prior to changeing that setting), although now there's a newer BIOS.

 

But I realized today that disabling the C-states is not really an option, the power consumption of the server vent up like a 100W when I disabled the C-states, I think from the attached image you can determine the time pretty accurate as to when I reboot with the C-states disabled.

 

I took a photo of the screen at one of the crashes as well if that could of any help I'll atttach that as well.

Again sounds like a board/BIOS problem, try the new BIOS.

 

 

On 12/20/2021 at 7:38 PM, JorgeB said:

Again sounds like a board/BIOS problem, try the new BIOS.

 

 

As an update to this thread:

 

I ran a memtest, only one pass, and it was OK.

 

So I went ahead and updated the BIOS again, and right now I am at 15 days uptime, which is a record, lets hope I didn't jinx it with this post

 

At the same time as I updated the BIOS I also changed a setting that disabled the system from booting from one of the disks attached to my HBA-card (this also cut boot-times as the HBA card does not do it's own boot-thing any more)

 

So I'm not sure wich of this actually was causing my issues, but hopefully this issue is solved now.