[plug] Ryzen

[Brad Campbell]

On 30/03/17 04:23, Richard Meyer wrote:

>> The latest it87 git tree has a workable hardware monitor module. AMD
>> still hasn't released the details to enable proper hardware monitoring,
>> but it's only a matter of time.
>>
>> If you buy an Asus motherboard with "Languard" on it, don't plug it into
>> a PoE enabled switch or you might : A) kill your switch, or B) let the
>> magic smoke out of the motherboard (I did both). But other than that,
>> just do it!
>
> That sounds like a talk in the making   :)

Nothing like spending $1500 on hardware and hearing that quiet "zap, 
phut" sound and getting a waft of the resulting magic smoke smell as you 
plug the network cable in before you realize your whole network is down 
and switch is dead.

Damn Asus and their ricer gamer crap. They're beginning to look like the 
Monster Cable of Motherboards.

Now, the Ryzen stress tested fine by itself for a couple of days (ie 
just sitting on a cardboard box with an SSD on it), but after migrating 
it into my staging system I found an issue.

The staging system has 2 LSI SAS cards and currently has 8 6TB drives in 
a RAID6. With a 4.10.6 kernel I was seeing soft-lockups under load 
resulting in rcu stall messages on the console and everything coming to 
a grinding halt. This took anywhere between about 30 minutes to 3 hours 
to trigger and only happened when under load (ie transferring files over 
bonded GigE interfaces or doing RAID checks or rebuilds).

I upgraded to 4.11-rc4 and it seems to have gone away, having survived a 
13 hour RAID check and 7 hours of networked file transfer, I'm going to 
press send on this and see if it crashes.

So if you are going to go down this route make sure you are using the 
latest possible kernel.

Regards,
Brad