Hello!

 

I've been trying to modify the "up_threshold" value of my ondemand governor, and I can't seem to get it working.

 

I have followed various guides/advice on this site as well as on the interweb, and for whatever reason, I cannot access up_threshold to modify the average cpu usage required to up the speed of my processors.

 

I have no problem swapping between scaling governors (I have the choice of performance or ondemand), I can list all the min/max values etc, but that is the extent of my ability. Also, I can confirm which governor is currently active with the following command:

 

cat /sys/devices/system/cpu/cpu0/cpufreq/scaling_governor

 

From what I can see, the problem is that even though the ondemand governor is working (I have confirmed that cpu speeds are jumping between 800mhz and 2200mhz, depending on load), there is no "ondemand" directory in: "/sys/devices/system/cpu/cpu*/cpufreq/". Because that directory does not seem to exist, I cannot change the up_threshold values to make the cpu scaling kick in earlier using the following command:

 

sudo sh -c "echo 20 > /sys/devices/system/cpu/cpu0/cpufreq/ondemand/up_threshold" 

 

Any ideas why I am having this problem?

 

Relevant info:

 

Unraid 5.0-rc6-r8168-test2

mobo: Gigabyte GA-MA790GP-UD4H

cpu: Athlon x4 600e

 

Here is a site/post I found particularly useful: http://www.pantz.org/software/cpufreq/usingcpufreqonlinux.html

 

 

Perhaps because the kernel was not compiled with that feature???

Perhaps because the kernel was not compiled with that feature???

 

Well that would certainly explain it! . Thanks for the response, Joe.

 

I had hoped because I could still select the ondemand or performance governor, that the ability to modify the up_threshold value was an option, and I was simply doing it wrong.

 

Oh well. I guess I'll just leave it set to performance for now.

 

For the curious, the reason I was hoping to modify that setting was that my server doesn't seem to kick up the speed of the processor unless under the heaviest of loads. Because of this, network transfers are significantly slower then when the processor is functioning at full speed. I can reliably recreate every time a difference of approximately 30 MB/s when transferring files from PC to cache drive (50MB/s vs 80 MB/s), when changing between ondemand and performance governor...

 

When under performance governor control, my 4 cores are running full out at 2200mhz. When using the ondemand governor, they generally stay at 800mhz...

 

I also came across some posts in ubuntu forums that state that the default up_threshold value was set to 31% before Jaunty, but now is set at 95%... Could something similar have happened in the UnRaid Kernel?

 

This leads me to extrapolate that perhaps this is one of the reasons my parity checks have been dog slow the last few versions... (NOTE: Haven't tested yet. Just a wild shot in the dark!!!)

 

I will definitely need to do some more test/reading/learning in this area, but thought I would share my thoughts