Power measurement of Arria 10 FPGA

[ADua0]

Hi,

I want to calculate the power consumption of fpga when I am running my opencl code on Arria 10 FPGA, is there any tool that can help me measure power

[HRZ]

What board are you using? Nallatech and Bittware Arria 10 boards come with on-board power sensors and they provide exampels in their documentation as to how the sensor can be read in standard C code. I think Intel's reference board also comes with some power measurement tool. For other boards, you should probably consult the board manufacturer to see if they provide a power sensor on the board. If not, you will be limited to either using a PCI-E riser cable and measuring power on that, which would be accurate but potentially dangerous (due to possibility of shorting during measurement), or use a Plug-In Power Meter and subtract the measured power consumption with and without the board, to get the board power usage, but this will be highly inaccurate.

[ADua0]

@HRZ​ , I am using ARRIA 10 GX FPGA board, I found one source of https://www.intel.com/content/www/us/en/programmable/support/support-resources/operation-and-testing/power/a10-power_estimator_download.html, to download power estimator, but I am not able to edit anything, it needs password to make changes . Do you know any other solution ?

[HRZ]

That is a power "estimator", not measurement tool. Maybe one of the Intel mods can point to a power measurement tool/method for the Arria 10 GX board.

[ADua0]

Oh okay I got, @HRZ​ , I used quartus_pow that did generated estimate power, should I trust this report generated from quartus_pow.

[ADua0]

Oh okay I got, @HRZ​ , I used quartus_pow that did generated estimate power, should I trust this report generated from quartus_pow.

[HRZ]

That is also one option, but quartus_pow assumes a signal activity of 12.5%. Depending on the switching characteristics of your design, the value you get from quartus_pow could be close to the real value, or very far from it. The static power usage should be relatively accurate; it is the dynamic power usage that is the main source of difference between the real power usage and the estimation you get from quartus_pow.