Anybody focused on the max power of the k20c?

Hi all,
Have you focused on the power of the k20c?
I used the SHOC benchmark to test the k20c. And the max power of the k20c is just about 150w, and it would not be bigger than 160W,even thoug the GPU utilization is reached 99%…
I have tied the SHOC benchmark on Quadro 4000/Tesla M2090/Grid K2, the card’s power could reach 95% of the TDP. So, I think the benchmark tool is no problem.
I just doubt something is wrong with my k20c.
By the way ,I have checked the Power limite is 225W.

*******************************************************************************
Mon Jul 15 19:43:51 2013       
+------------------------------------------------------+                       
| NVIDIA-SMI 5.319.32   Driver Version: 319.32         |                       
|-------------------------------+----------------------+----------------------+
| GPU  Name        Persistence-M| Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |
|===============================+======================+======================|
|   0  Tesla K20c          Off  | 0000:06:00.0     Off |                  Off |
| 35%   47C    P0   146W / 225W |      283MB /  5119MB |     99%      Default |
+-------------------------------+----------------------+----------------------+
                                                                               
+-----------------------------------------------------------------------------+
| Compute processes:                                               GPU Memory |
|  GPU       PID  Process name                                     Usage      |
|=============================================================================|
|    0      6316  ./SGEMM                                              267MB  |
+-----------------------------------------------------------------------------+
****************************************************************************************

I just doubt something is wrong with my k20c.
By the way ,I have checked the Power limite is 225W.

==============NVSMI LOG==============

Timestamp                           : Mon Jul 15 19:56:42 2013
Driver Version                      : 319.32

Attached GPUs                       : 1
GPU 0000:06:00.0
    Power Readings
        Power Management            : Supported
        Power Draw                  : 149.76 W
        Power Limit                 : 225.00 W
        Default Power Limit         : 225.00 W
        Enforced Power Limit        : 225.00 W
        Min Power Limit             : 150.00 W
        Max Power Limit             : 225.00 W

Could anybody help me to confirm this problem?
or could anyone recommend one test tool?
Thank you.

To go closer to the power limit, just use DGEMM.
SHOC is not a good benchmark for power.

Currently,the SHOC has been updated,and add a DGEMM script.
And I have tried it on k20c, but the power is still about 150W.
Maybe, the SHOC is not good enough for power, though it semms good on other architecture’s GPU.

Could you recommend one test tool, or could I download one on the nvidia website?

Thank you.

I am not sure why relatively low power consumption presents a problem, normally that is considered a good thing. To check out the basic capabilities of your GPU you could measure the performance of some large, square, DGEMMs. As a point of comparison, I observe roughly 970 - 1100 GFLOPS on my K20c here, depending on which of the four transpose mode combinations I run DGEMM with. This is with CUDA 5.5, I have ECC enabled.

Execution of these large DGEMMs (I am using m = n = k = 8192) results in a power draw of between 160W and 170W on this K20c as reported by nvidia-smi, so that seems roughly consistent with your numbers. I notice that power draw differs depending on the transpose mode combination, so you might want to try all four combinations. There is probably some amount of variation between invidual GPUs, both in terms of the actual power draw and the noise level of the sensors that nvidia-smi queries. You might want to run a loop doing DGEMMs for a couple of minutes until you see steady numbers reported by nvidia-smi.