Power Capping by Power Governor and Power Gadget on Linux

[Kevin]

Hi everyone,

When I tried to use Power Governor(https://software.intel.com/en-us/articles/intel-power-governor) and Power Gadget(https://software.intel.com/en-us/articles/intel-power-gadget-20) to get the power consumption info of CPU, I got the following error:

RAPL not supported, or machine model 306f2 not recognized.

Init failed!

When I followed the README.md to "modprobe msr" and "modprobe cpuid", it shows:

FATAL: Module mrs not found.

FATAL: Module cpuid not found.

I found any posts in the forum but didn't find any problems like this.

The Purpose is to get the power consumption/temp info in a program.

Does anyone have any ideas how to deal with the problems above of these two tools or other methods to get the power consumption in a program? 

Best wishes,

Jiawen

[Kevin]

Hi everyone,

Does anyone know how to figure out this problem? I appreciate if anyone could give me any suggestions or ideas.

Thanks,

Jiawen

[Carlos_P_]

Hi,

What's your CPU model?

Regards

[Kevin]

Hi Carlos,

The CPU is  Eighteen-Core/36-Threads Intel® Xeon® E5-2699 v3 2.3GHz-45MB Shared L3 Cache-9.6GT/s QPI-145W-22nm. Does it support the rapl architecture?

Carlos P. wrote:

Hi,

What's your CPU model?

Regards

[Carlos_P_ · Accepted solution]

With that details, I believe your "DisplayFamily_DisplayModel" is 06_3FH.
Well, just checked on the Intel® 64 and IA-32 Architectures Software Developer’s Manual Documentation by that ID and I find RAPL registers available... 
Try running this sample code: http://web.eece.maine.edu/~vweaver/projects/rapl/ (Userspace Access without perf)

Regards

[Kevin]

Hi Carlos,

When I compile the code, I still get the error message as below:

[liu@fornax ~]$ gcc rapl-read.c 

/tmp/ccKurDYv.o: In function `rapl_msr':

rapl-read.c:(.text+0x3f6): undefined reference to `pow'

rapl-read.c:(.text+0x43b): undefined reference to `pow'

rapl-read.c:(.text+0x480): undefined reference to `pow'

collect2: error: ld returned 1 exit status

Do you know what is the reason?

Thanks,

Jiawen

[Carlos_P_]

Hey,

you need to compile with the flag -lm

regards

[Kevin]

Hi Carlos,

After I run it, I got the following info showing that No perf_event rapl support found. I still didn't get the power consumption of CPUs. Do you know what's the problem?

[liu@fornax ~]$ sudo ./a.out 

No perf_event rapl support found (requires Linux 3.14)

Falling back to raw msr support

Found Haswell-EP CPU

Checking core #0

Power units = 0.125W

CPU Energy units = 0.00006104J

DRAM Energy units = 0.00001526J

Time units = 0.00097656s

Package thermal spec: 145.000W

Package minimum power: 69.000W

Package maximum power: 290.000W

Package maximum time window: 0.045898s

Package power limits are unlocked

Package power limit #1: 145.000W for 0.009766s (enabled, clamped)

Package power limit #2: 174.000W for 0.002930s (enabled, clamped)

Package energy before: 75131.813354J

PowerPlane0 (core) for core 0 energy before: 0.000000J

rdmsr:pread: Input/output error

Thanks,

Jiawen

[Carlos_P_]

Hi Jiawen,

https://software.intel.com/pt-br/forums/software-tuning-performance-optimization-platform-monitoring/topic/557622

Check that topic best reply from Patrick. Maybe you have a similar problem so, there are some MSR's that are not compatible with your processor, maybe they are the same since it is a E5 too.

Check if it will help you!

Try it and give some feedback :)

Regards! 
 

[Koji_T_]

Hi all,

I'm having the same problem. The CPU is Xeon(R) CPU E5-2680 v3. Isn't there a way to make it work?

Thanks.