Changes That You Won't Find In Linux 4.13

[starshipeleven]

LOL, are you really this PR blinded? Becachefs is in it‘s baby days.

I think you should check your sarcasm detection system.
Mine showed obvious sarcasm in his post.

[Zan Lynx]

Installing the it87 module from https://github.com/groeck/it87 worked for my Ryzen. The temp and fan is good but voltages are all off.

[ferry]

Btrfs: Hot data tracking and moving to faster devices

It would be so useful to just throw in 2 ssd's next to our 4 hdd's in our raid10 system and instantly get ssd speed backed by hdd storage space. This has been marked as 'in development' since for ever.

[starshipeleven]

Many servers I've seen seem to only have 3 fan speeds: full, low, and off. When it comes to servers, if you have temperature issues, you've got something else bigger to worry about.

with the "low" setting still moving enough air to achieve a VTOL takeoff and require hearing protection, let's not forget that.

[waxhead]

Btrfs: Hot data tracking and moving to faster devices

It would be so useful to just throw in 2 ssd's next to our 4 hdd's in our raid10 system and instantly get ssd speed backed by hdd storage space. This has been marked as 'in development' since for ever.

This feature was dropped and was supposed to be part of the VFS instead. I proposed what I call "supercache" a while ago that I think would be far superior. Imagine that for regular harddrives for example, btrfs could transparently mirror hot data in a raid0 cache over the fastest group of disks. If the raid0 mirror breaks there is still a copy on a more redundant (but slower) configuration. Might be just as nice especially if you have lots of disks and low throughout.

[caligula]

RyZen is at least operating pretty well for basic functionality in Linux but it's a little disappointing that they can't even get simple temperature monitoring working.

Just compare the AMD's R&D budget with Intel's.. as Ryzen users here know, you already CAN monitor the CPU temp, but the sensor is beneath the chip on the mobo. As for the accuracy, it might be off by few degrees, but it's not a big deal at this point. If you're curious about Ryzen's stability, it is stable. If you're planning to overclock, use mprime or ffmpeg to make it sweat for few days, then decide. My Ryzen 1700 goes up to 3,8 GHz easily. 3,9 starts to fail after 20 minutes of ffmpeg transcoding. I haven't touched the voltages that much though, I've read there's still some headroom left.

The motherboard sensor seems pretty accurate, at least in my case. It may be off by a few degrees but accuracy is somewhat irrelevant as long as the cooling system works properly. In other words, as long as the motherboard can monitor and compensate the direct CPU temperature, that's all that matters.

Right.

[Space Heater]

Have there been any movements to upstream support for temperature monitoring on ryzen?

[caligula]

Have there been any movements to upstream support for temperature monitoring on ryzen?

The specs aren't out yet AFAIK https://github.com/groeck/lm-sensors/issues/16

[k1e0x]

you won´t monitor the many HPC- or serversystems temperatures with gui- or commandline-tools, do you? I would think that those systems would ring a bell or throttle down automatically due to settings in the bios/efi.
it would only be useful for overclocking or personal workstations that crash in summer months. but the enterprises / HPCs I know keep their servers in special rooms which are temperature monitored as a whole.

No they don't ring a bell usually. heh. They throw an alarm to one of the various monitoring utilities out there, but those utilities need to be able to pick up the sensor data. And.. Yes, enterprise will want that for Epyc.

I think FreeBSD has some hacks to get this working but the problem is AMD hasn't released the documentation for how the sensor works in Ryzen.. from its design it looks like the mother of all temp sensors to control its XFR so there may be some issue here.

[Azrael5]

bus1 will be another failure after kdbus.