[kondbg]

I also built a Ryzen machine for development. It's great when it works, but I've found that Ryzen is unstable on Linux (Ubuntu 16.04). Every once in a while, I get kernel errors like

   NMI watchdog: BUG: soft lockup - CPU#9 stuck for 23s!

which requires a hard reset. This behavior doesn't occur on Windows, though, so if you use Windows for development, you should be good.

[Pengwin]

Did you try a later kernel? out of the box ubuntu 16.04 is on the longterm 4.4 kernel. It looks like some ryzen features and patches have been added to 4.10, and they probably were not back-ported to the longterm kernels.

[look_lookatme]

Not OP, but just built at Ryzen 1600 box. I've had _more_ instability when running 17.04 and settled on 16.04 which has been mostly fine, but hard crashes occasionally.

[bdcravens]

> hard crashes occasionally

This sounds disturbingly unacceptable yet accepted

[look_lookatme]

I custom built a machine last week for the first time in over a decade and from a fairly new processor and a just released video card. If I had more confidence in my PC building I guess I'd be more upset, but there are a lot of variables here and I'm still in a honeymoon phase.

I installed the official AMD RX580 drivers and it's been stable since.

[dom0]

Nothing out of the ordinary with a new platform. While Linux often takes longer to work this stuff out, Windows often has had similar issues in the past as well.

[shady-lady]

Not sure if intent but this is AMD issue.

Having experienced first hand AMD driver instability on both Windows & Linux, AMD have lost my custom for the next 10 years..

Multiple re-occurring driver crashes using their main graphics card product line(RX380) on Windows 10.. So, pretty mainstream and yet having driver crashing (even when doing non-intensive tasks e.g. web browsing)

For the record, I'm not so sure Nvidia is any more stable either. The only (constantly) stable graphics provider over the years has been Intel's on-board graphics.

[pawadu]

You can use 16.04 with a newer kernel. There is even a deb package for 4.8

[fulafel]

The official 16.04 hwe "edge" kernel is currently at 4.10.x: http://packages.ubuntu.com/xenial/kernel/linux-generic-hwe-1...

[kondbg]

Yes, that log entry was from running 4.10.11.

[clarry]

I recently built a system with R7 1800X and Arch is randomly resetting when running its 4.10.* kernels. Fedora, with 4.11 rc builds (the real thing is out now btw) has been rock solid with weeks of uptime, running games & browsers & development stuff.

[noir_lord]

Good to know Fedora is worth a try, I'm a Xubuntu user and have Ryzen parts coming, won't need it til June so I'm hoping Ubuntu gets it sorted before then but if not I can live with Fedora for a while.

[luca_ing]

I experience the same issue on my Skylake i5 :-)

IIRC in my case it goes away once I install bumblebee. Apparently something to do with switching between internal Intel graphics and the dedicated Nvidia chip, and not at all related to your issue, except for the symptom.

It's comedic because after bootup I have about 90s until lockup, so I have to be lightening-fast to type the commands to install bumblebee. If I'm too slow: reset machine, try again.

FWIW, this is on a notebook, running Ubuntu 16.04.

[mixmastamyk]

Try adding it to a startup script.

[luca_ing]

Thanks for the suggestion.

I was unclear, this only happens after a fresh install, before I've installed Bumblebee and its dependencies. Once that is out of the way, I don't need to worry about it any more.

[look_lookatme]

I also have been experiencing lock ups with my Ryzen system. I thought it was my RX580, but now I'm thinking it might be the CPU/mobo. Did you go with a B350?

[kondbg]

Yes, MSI B350 Tomahawk

[look_lookatme]

Hrm... same here.

[FnuGk]

Open a bug report?

[paulmd]

You might want to try disabling SMT and see if that helps. No promises, could be many other things, but it might be worth a shot.

SMT support on Ryzen is really flakey, it's almost always a wash and often can hurt performance and I wouldn't be surprised to see it cause this kind of bug as well (especially in the early days of kernel support).