This article is more than 1 year old
IBM's beefy Power6 boxes slaughter Java
While low-end Dellintel trumps IBM's 4.7GHz heater
Pop quiz. You're in desperate need of making Java fly and have the coin to prove it. Do you shell out on IBM's new 4.7GHz Power6 eye candy or keep on replacing Unix kit with x86 systems?
For the moment, it looks like you will want to saunter into the 4GHz realm if you're a midrange server type. IBM has just released SpecJbb benchmark results for its Power6-based p 570 servers. The figures show IBM assaulting both the x86 and RISC crowd with four-socket and above systems.
IBM, for example, notches a score of 346,742 bops with a four-socket, eight-core server running 16 software threads. That beat out the score of 217,334 bops set by a similarly configured Xeon-based ProLiant system from HP and a score of 158,174 set by an Itanium-based HP server.
On the eight-socket front, IBM beat out competing Unix systems from Fujitsu, Sun Microsystems and SGI, along with a Xeon-based Fujitsu box.
IBM, however, did not do so well against lower-end x86-based systems. In fact, it still trails one-socket and two-socket Xeon-based systems from Dell even with the 4.7GHz heft.
The low-end struggles could soon turn into high-end struggles, says David Kanter of Real World Technologies, which has a nice summary of the benchmarks results.
"(Those results) have to be somewhat problematic, since Intel is going to be coming out with a XeonMP version of Clovertown kind of soon," Kanter told us.
Of course, if you're an AIX fan, you might not care what Intel s(c)hips this year or the next.
Overall, these Java benchmarks bring us a bit closer to understanding Power6's true performance. Lord knows that you won't get very far depending on TPC scores where the vendors have shelled out millions of dollars and thousands of hours to beat each other up over pennies per transaction. ®
Bootnote
Before the hate mail comes rushing in, we'll note that reader RG posting in the comments section makes a great point about IBM's linear scaling with regard to bops/JVM. IBM also uses half as many cores as Dell on the one- and two-socket tests.