tl;dr: Using nbench today is wrong. The S500 suffers from thermal issues preventing 'full load' operation over longer periods of time.
First of all nbench is really outdated and results vary massively depending on compiler flags (you get easily 10% difference or more due to different CFLAGS settings. Therefore useless to compare different nbench results found on the net)
More importantly nbench is single threaded that means it runs on only one CPU core (nbench is 20 years old, now we have SMP everywhere and for a system with more than 1 CPU core things start to get difficult. You can not directly interpolate the single thread performance to the performance the system is able to achieve when running on all cores in parallel). But the most important drawback of using an inappropriate benchmark like nbench on modern architectures is that this single threaded behaviour also massively influences the performance results you get.
On x86 for example there exist the 'turbo boost' concept. If only one or 2 cores are under full load the clock speed of these cores will be increased way more than if all CPU cores would be active. So no way to get an idea how performant a multi core system would behave under full load when using a tool like nbench (that has been written back at a time when CPUs were really different!).
And the very same problem we have with the S500 (single CPU performance looks better than when running on all CPU cores) but in a slightly different way. The S500 tends to overheat after several minutes under full load (based on my tests full temperature will be reached after 10-12 minutes under full load). And then throttling should jump in and performance decreases drastically. But using nbench which only utilises a single CPU core the load is too low to get even close to 'full load behaviour'.
I would doubt that it's possible to run the S500 at 1.1 GHz without a very effective heatsink or a fan over longer periods of time without the CPU cores starting to throttle down. Have a look at the wiki (containing an external link at the end!) for details: http://wiki.linux-xapple.org/w/index.php/CPU#How_to_check_the_SoC.27s_temperature
BTW: If you want a performant ARM system in 2015 then don't think about Cortex-A9 but A-53, A-57 or A-72 instead.