Watts (chipotle) wrote,

PowerPC 970 benchmarks

A (possibly) interesting thing I noticed poking around: one of the main criticisms of Apple's recent G5 benchmarks is, essentially, that you should always use the fastest scores for a given processor, not just go off and do your own and trumpet those results.

The interesting thing is from this PDF article from the October 28, 2002 issue of Microprocessor Report, which quotes IBM's estimated SPECint2000 and SPECfp2000 scores for the PPC970 at 937 and 1051 respectively. And that's the 1.8 GHz part, not the 2.0 GHz part; if we extrapolate those numbers to 2 GHz, we get SPECint2000 1041 and SPECfp2000 1167.

Since we're using IBM's scores instead of Apple's for that side, we'll use Intel's scores instead of Dell's for its side. Intel's scores for the 3.06 GHz P4 are SPECint2000 1088 and SPECfp2000 1077.

Which would actually put them both in about the same league, with neither side able to claim "our processor stomps your processor." So everybody put a sock in it.

(For the record, in case someone asks, AMD's scores for the single-processor Opteron 144, their fastest single processor part, are SPECint2000 1170 and SPECfp2000 1219. AMD quotes peak rates, while the others are quoted in base rates.)

  • A quick note

    Posts on Coyote Tracks are supposed to be cross-posted here, but it's clear the cross-poster isn't, er, posting. I apologize. I'll look into it,…

  • A better Amaretto Sour

    I’m pretty sure I was introduced to the amaretto sour in college by my roommate’s girlfriend. I liked it—because I like amaretto—but I…

  • Cotton, hay, and rags: giving bias the veneer of rationality

    As you’ve surely heard by now, a mid-level engineer at Google—he’s anonymous, so I’ll call him Mr. Rationalface—wrote a memo called…

  • Post a new comment


    Anonymous comments are disabled in this journal

    default userpic

    Your reply will be screened

    Your IP address will be recorded