Andrew Mobbs (mobbsy) wrote,
Andrew Mobbs

Dear Sun Microsystems, it's really not helpful when you compare a dual-core US IV+ to a single core of a POWER5 chip. I fully believe you when you say it's system throughput that matters, not per-thread performance, but that doesn't help when Oracle sell licence on a per-core basis.

Actually, Dear World, can we settle on talking about "cores", or will increasingly complex SMT/CMT implementations make even that too hard?

Or, Dear Oracle, please licence your database software in some sensible manner, rather than 1*cores for single core chips, 0.75*cores for multi-core other than AMD/Intel and Niagara, 0.5*cores for AMD/Intel multi-core and 0.25*core for Sun Niagara systems. OK, so this is still fractionally more sensible than the old Oracle 8 "power unit" pricing, where faster CPUs of the same model cost more, but still, I mean, like, guys...

Oh, and word to the wise, if you're ever talking to Sun salespeople about Ultrasparc performance, ask about the "mvalues" compared to Opteron systems. That's Sun's internal comparison metric. It's more enlightening than trying to argue with the salespeople about per-core performance based only on your own careful analysis of published benchmarks.

  • (no subject)

    Last week I poured the cremated remains of my father into a river. From there, that material will flow through the town he lived in, into the sea,…

  • Moving house!

    We're moving house soon… details to follow in a less public post, or email me. However, we're getting rid of some bits and pieces of…

  • (no subject)

    Yesterday, I made sausages. This was sufficiently exciting to cause me to actually write something on LJ for once. One of our wedding gifts was a…

  • Post a new comment


    default userpic

    Your IP address will be recorded 

    When you submit the form an invisible reCAPTCHA check will be performed.
    You must follow the Privacy Policy and Google Terms of use.
  • 1 comment