04-27-2018 06:33 AM
We have 2 custom boards containing Virtex-7 690 FPGAs.
On one of the boards, commercial grade Virtex-7 690 is mounted and on the other one industrial grade Virtex -7 690 is mounted.
When only the power of 1V core voltage is supplied to the boards we observe significant power consumption differences between the boards.
We observe that the board with industrial grade FPGA consumes 40% more power with respect to the board with commercial grade FPGA.
Is this normal? We do not expect different power consumptions for commercial and industrial grade FPGAs.
When both of the FPGAs on the boards are configured with the same configurations there is no difference in terms of functionality of the FPGAs. However, the power consumption difference between the boards tends to increase.
Any suggestions to diagnose defects on the boards?
04-27-2018 06:38 AM
Can you tell us how much power is consumed in each?
Can you tell us what the power consumption is when you have both devices powered but not configured?
Are they in the same package?
What is report_power or XPE telling you?
05-16-2018 06:04 AM - edited 05-16-2018 06:06 AM
both commercial and industrial FPGAs are in the same package.
When commercial grade FPGA is powered in JTAG mode and no configuration is loaded, the board (with FPGA and other componenets) consumes 3 Watt power.
When industrial grade FPGA is powered in JTAG mode and no configuration is loaded, the board(with FPGA and other componenets) consumes 4 Watt power.
In order to test power consumption while FPGA is running we generate a simple configuration.
According to (report_power) this configuration consumes 0.211 Watt from 1V power supplying Vccint and Vccbram.
When the power consumed by this configuration is measured for 1V power supplies on the boards we observe 0.26Watt power consumption for commercial FPGA and 0.53Watt power consumption for industrial grade FPGA.
is this normal?
05-16-2018 10:39 AM
when you do report power, export the data for XPE also.
Then open the XPE sheet and import the design
When I do this you can change the temp grade and I see with my simple design the static power at the worst case is the main thing affected by the temperature grade.
In this case we have a really simple design and I see it has only about 20mW difference in static power given with I grade being lower.
05-16-2018 12:14 PM
There's a simple physical explanation why Industrial grade parts will consume more static power. When the silicon process is adjusted for extended temperature ranges, you need to boost the speed of the silicon to handle the higher temps.. and this requires heavier doping, which makes the silicon leak more.. it's a tradeoff. Commerical temp silicon can use less doping, which makes it leak less.
Take your pick. Cheap, Fast, Low Leakage.. but you only get 2.
05-16-2018 03:05 PM
05-16-2018 03:23 PM
yes, it's likely that the parts are not really "adjusted" for commercial vs industrial.. but the parts that get binned for industrial temp ranges are likely to be the faster and leakier parts.. This is why there aren't any -3 parts with Industrial temperature.. it's too far out on the edge.