Update:

>> Just for the purpose of overview I compiled a list of the DMIPS/MHz per core
>> figures for various PPCs in ascending order (AltiVec capability is generally
>> ignored, obviously):
>> [...]
>> e5500: 3.0

> The table on page #2 of the QorIQ AMP fact sheet PDF indicates that
> the e6500 core didn't improve in that regard (comparing e6500's per-thread
> performance to e5500's per-core performance).

The QorIQ AMP T4240/T4160 fact sheet claims "6.0 DMIPS/MHz per core" and at the same time for dual threads "1.7 times the performance of a single thread". To me this sounds strange as I think it should be either still 3.0 DMIPS per MHz and core or about 5 DMIPS per MHz and core at best in case of dual-threaded execution. Or does the Dhrystone benchmark really scale linearly with the amount of executable threads?