Macintouch has an iMac G5

Posted:
in Current Mac Hardware edited January 2014
They've posted a preliminary Xbench shootout against an iMac G5 1.8 and some comments on the system. Expect they'll have something more in depth soon. Might do well to watch for their reports.



Ankly



Macintouch



edit: Would some kind mod fix this poor tired Mac users idiotic mistake with the thread title? it should be iMac Intel





Yeesh!

Comments

  • Reply 1 of 5
    I'm I alone in thinking the Xbench results are extremely ambiguous?



    Taking Ric's Xbench scores, with my (facetious) comments added:



    iMac Core Duo 2.0GHz vs old iMac G5 1.8GHz:



    CPU Testt76.6 (core duo) vs 87.55 (old iMac G5)

    - Possibly true, but it makes that Core Duo processor seem pretty pathetic



    Thread Testt198.3 (core duo) vs 49.75 (old iMac G5)

    - but it's a whopping 4x as good at threading?



    Memory Testt103.6 (core duo) vs 74.5 (old iMac G5)

    - an unexpectedly huge benefit from increasing to a 667MHz FSB from 533MHz



    Quartz Graphics Test: 68.6 (core duo) vs 81.2 (old iMac G5)

    - That fancy ATI X1600 isn't helping with the graphics is it, it's worse than the old iMac's nVidia 5200.



    OpenGL Graphics Testt140.4 (core duo) vs 90.4 (old iMac G5)

    - But excels at OpenGL graphics.



    User Interface Testt18.6 (core duo) vs 64.3 (old iMac G5)

    - What the hell UI is that measuring, because whatever it is the Core Duo isn't doing very well.



    Disk Test: 90.6 (core duo) vs 41.0 (old iMac G5)

    - A disk test should be testing, oh I dunno, disk performance, so such a huge performance increase would suggest a VERY fast drive, but then later Ric says he copied a 207.5MB file in 10 secs on the Core Duo which took 12.7 seconds on the old iMac. That real world test is pretty much in line with expectations (slightly faster, newer drive), so just what disk performance is Xbench measuring?





    Based on these cursory observations of the numbers, something about Xbench seems very fishy. Especially given anecdotal reports of how snappy programs like Safari seem, which Xbench would suggest would not be the case given abysmal CPU, UI, and Quartz performance.



    Xbench may be measuring something, but whatever it is doesn't seem to be very helpful...
  • Reply 2 of 5
    chuckerchucker Posts: 5,089member
    Quote:

    Originally posted by michaelb

    I'm I alone in thinking the Xbench results are extremely ambiguous?



    No. Xbench results are, have been and always will be completely useless. You can make them more useful by doing multiple runs, averaging them, and praying the medium deviation is reasonably small, but usually, it won't be.
  • Reply 3 of 5
    Quote:

    Originally posted by Chucker

    No. Xbench results are, have been and always will be completely useless. You can make them more useful by doing multiple runs, averaging them, and praying the medium deviation is reasonably small, but usually, it won't be.



    Ric Ford averaged out the scores of four different run-throughs after restarts, according to the MacInTouch site.
  • Reply 4 of 5
    pbpb Posts: 4,233member
    Quote:

    Originally posted by LudwigVan

    Ric Ford averaged out the scores of four different run-throughs after restarts, according to the MacInTouch site.



    It matters not. We talk about a completely new architecture here and although Xbench runs natively in the new iMacs, I have serious doubts it has been tweaked to run correctly on Intel.



    However, there is no other extensive benchmark tool for the Mac. This is perhaps the reason why people still bother with this POJ. Now that every Mac has iLife on it, someone should write a test script to put some serious stress on the machine, by calling the iLife applications, and get more meaningful numbers for the same OS and iLife version.
  • Reply 5 of 5
    costiquecostique Posts: 1,084member
    Quote:

    Originally posted by michaelb

    I'm I alone in thinking the Xbench results are extremely ambiguous?

    Thread Testt198.3 (core duo) vs 49.75 (old iMac G5)

    - but it's a whopping 4x as good at threading?



    Yes, it looks plausible, given two CPUs vs one and the fact that the kernel should run faster on x86 than on PPC (at least marginally), all things being equal. In real-world apps I'd expect something closer to 2.5x, because Xbench threads are most likely dummy ones and easily fit into the big shared cache.

    Quote:

    Memory Testt103.6 (core duo) vs 74.5 (old iMac G5)

    - an unexpectedly huge benefit from increasing to a 667MHz FSB from 533MHz



    As fas as I can tell, a more efficient memory controller greatly contributes to this. Remember that G5 suffers from huge memory access latencies, like 9 CPU cycles or something like that.

    Quote:

    Quartz Graphics Test: 68.6 (core duo) vs 81.2 (old iMac G5)

    - That fancy ATI X1600 isn't helping with the graphics is it, it's worse than the old iMac's nVidia 5200.



    This, well, does seem fishy. Something must be very wrong with the test. I bet the ratio must be close to that of the OpenGL test.



    The UI and disk tests are absurd. Yes, the old iMac did not have SATA (or did it? - Edit: it did indeed). Yes, the x86 drivers ought to be faster, but I cannot believe by so much. Because even if the disk system received a 30% boost, I WANT A NEW MAC, Intel be damned.
Sign In or Register to comment.