Jesper Frimann (jesperfrimann@gmail.com) on 2/16/10 wrote:
---------------------------
>One example I noticed is this one:
>http://www.ideasinternational.com/benchmark/ben020.aspx?b=f88a4f62-6cdc-40b9-abaf-a4b481d109e5
>
>POWER 7 1 Chip 6 Cores@3.3GHz -> 257.143 records/hour.
>Nehalem 2 Chip 8 Cores@2.93GHz -> 229.885 records/hour.
>
>Which means:
>POWER 7 -> 42.857 records/hour per core.
>Nehalem -> 28.736 records/hour per core.
>
>// Jesper

Hmmm, does the fact that having 250K empoloyees for the Nehalem result versus 10K employees for POWER7 make a difference? I don't know but this seems odd? I can't tell if working through the larger dataset has any impact...

Also, I dug up the detailed report for these benchmarks from Oracle's site and the results seem questionable all around, but especially the POWER7 750 result appears to have a lot of guesstimates and "linear extrapolations" going on it in workload results table. This crude analysis on my part is based on the number of results that have what appears to be un-measured numbers. Heavy rounding going on whereas other results have none?

There are only two workload measurements in common that don't appear to be guesstimates. In both cases the Nehalem edges out the POWER7.

10,000 Employees (POWER7 - 750)
Workload Threads Hrly. Througput
Payroll Processing 12 1,161,290
External Archive 12 423,529

250,000 Employees (Xeon 5570 - DL580)
Workload Threads Hrly. Througput
Payroll Processing 8 1,395,349
External Archive 8 468,750

Take a look and let me know if you agree? I don't know this benchmark well enough to know how much fudging actually goes on. I'd be keen to see TPC results before I come to any conclusion on which CPU is the champ in synthetic benchmarks.

http://www.oracle.com/apps_benchmark/html/results.html