New on LowEndTalk? Please Register and read our Community Rules.
All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.
All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.
Comments
https://lowendtalk.com/discussion/comment/4008738
It's actually crazy how common unstable 7950X3D servers are.
I recently got a bunch of 7950X3D VPS's, and many of them have had stability issues, random reboots, freezes and node hardware failures.
I got 2 dedicated 7950X3D servers. The motherboard failed on one after like a month, and the other one has been randomly hard-freezing until manually cycling the power via IPMI.
All these servers are from different providers.
Are these boards are Asrock?
Seems they having issues on 7950x as well.
I hate these servers. They are riddled with issues. ASRock in particular. I've had over 10 board die in a few months in various datacenters. Apparently this is a specific issue with earlier models of their B650D4U.
7950XD itself has issues with nested virtualization / host passthrough. Certain applications simply kills the host system forcing a restart. Disabling that has resolved the majority of my issues.
The one that failed and was replaced is Asrock. The one that is randomly freezing has a Supermicro board.
No idea what boards all the VPS nodes have, but there is a clearly a pattern with Ryzen 9 servers. Not even only the 7950x. Had some other Ryzen 9's with bad motherboards causing random reboots recently too.
I've seen more hardware issues over the last few months since I started getting more Ryzen servers than the previous 20 years combined.
It's a strange issue for sure, but happens across all boards Supermicro or ASRock on the 7000 series (both X3D and non-X3D CPUs) and to an extent on the 5950X as well at times, sometimes more/less random depending on what CPU instructions the VMs are using.
While ASRock straight up doesn't seem to like the X3D CPUs -- I suspect the voltage issues which the X3D CPUs had at launch time with all Ryzen motherboards still haven't been completely fixed on the ASRock boards given the number of hosts that are having issues with that CPU<>Motherboard combination.
Looking at the random crashes,
host-passthrough which enables SVM (nested-virtualization) for VMs too, if in-use will cause issues on ASRock boards instantly, while Supermicro seems to bear it a bit more depending on what kernel version you are on.
It's a strange issue as it leaves no log either, hence feels like a hardware bug (due to the instant "reset like reboot") which happens each time this fault occurs.
Thankfully there's some (possible/usable) mitigations like disabling host-passthrough or not passing the svm flag and more people coming together and discussing the same for this,
host-passthrough which enables SVM (nested-virtualization) for VMs too, if in-use will cause issues on ASRock boards instantly, while Supermicro seems to bear it a bit more depending on what kernel version you are on.
The moment we disabled nested virt, the issues are gone
So we are turning the nested-vir off and will update you with the results.
How about when we disabled SVM from the BIOS. The VM's couldn't be loaded or created. So we had to enabled it back.
Thanks