Well, one of the things you have to consider in any 'study' like this is the configurations that were used....were they straight out of the box, turn the system on and test, or were any attempts made to change how the system works (to either speed it up or secure it). Either way, in alot of comparisions I have seen, you get more of a comparison of how the systems were setup than of the actual performance of the system ... a good example would be that in Solaris version < 8, there were no default tweaks of the TCP/IP stack... and a web server (Apache) running on a default configuration would literally choke under minimal load...a few stack changes later...the performance was orders of magnitude better...Sun's attitude at the time was we prefer to leave it to the system admin to configure the system how they wish...some of which Sun has changed in its later versions...

Regardless, at least with the Solaris systems...alot of the tests I saw at the time were not really reflective of the server's capability...

Can't specifically back up the numbers there, but did want to mention some related things I have personally seen.