Hi all,

my research group is doing research in power/energy/thermal management in MPSoC systems. What we would like to investigate is the impact that hardware and software design choices have on the power/energy/thermal profile of the system. Actually, we would like to investigate the effectiveness of different approaches w.r.t. demperature distribution across the chip, considering a multi-core system.
For our purposes, we are evaluating the possibility to buy an FPGA-based board to do our experiments, but I have a doubt on FPGAs... In fact, I am not sure that FPGAs (due to their technology) are well suited for such research, and let me try to explain you my doubt: my feeling is that FPGAs will not enable us to see the temporal differences in power consumption when a core is performing useful work and when a core has no workload because FPGAs are "always on". Am I correct with this?

Thanks for your help
Best regards

Simone

 

kingslayer
What is your definition of "always on"? Is that just on and sitting idle or is it on and taking inputs in, doing something with those inputs, and then outputting a value or set of values?

FPGAs are very well suited to your research... so much so that Xilinx went to the trouble to develop their power calculator specifically to help engineers understand exactly how much power the device in question would require based on resource usage. That is the key to power requirements for programmable logic.

As far as measuring temperature differences you use pyrometer while the device in question is operating or even get some temp sensors and place them on and around the part and take some measurements.

Good luck

 

Dear ghillie,
thanks for your reply.

With "always on" I mean that the device is just on and sitting idle.

For temperature sensors, I already studied the problem (FPGAs do not have on-chip sensors) so that it would be possible to synthesize thermal sensors through ring-oscillators.

However, thanks again for your reply. That was of much interest to me ;-)