I noticed that on a small number of machines I occasionally see a very high, obviously incorrect value reported for the CPU power (over 260000) . After some investigation, I think this is caused by the handling of rapl/dram values when wrapping. That is, it doesn't handle the case where only the DRAM value has wrapped (not the psys or pkg values); it will erroneously apply either psys_max_energy_range or pkg_max_energy_range, and then (correctly) dram_max_energy_range. On reflection I think a clearer way of handling this would be to handle psys/pkg/dram wrapping separately, rather than using 'total_energy'. I can make a pull request, though I think you're intending to refactor this code anyway.