Reduced is slower -- much slower

Discussion in 'macOS' started by bzgnyc, Apr 1, 2006.

  1. bzgnyc macrumors newbie

    Joined:
    Oct 19, 2005
    #1
    Hi,

    This turned out to be longer than I expected so let me break it down into several parts:

    1) Does your iBook or PowerBook feel much slower when operating in Processor: Reduced mode compared to Highest mode? Have you measured the difference? In either case, what release of the OS are/were you running?

    I have noticed a big slowdown and I am curious if anyone else noticed this on their iBook or PowerBook? I am trying to figure out if this problem is limited to this machine, this revision, this model line, this release of the OS, or some combination.


    2) What I mean by big slow down is factors of 2.5 or more in UI performance. Here are the UI results from different runs of XBench 1.2 on my PowerBook Al 1.25GHz (RevA, etc):

    Highest, Freshly booted/clean: 170-190
    Reduced, Freshly booted/clean: 65-75 (>2.5x slower)
    Highest, Browser/Dashboard: 110-122
    Reduced, Browser/Dashboard: 36-37 (>3x slower)
    [Note: I ran XBench at least three times for each case]


    3) Just to explain why this isn't just a numerical curiosity, let my explain what motivated all this:

    I was been noticing that my PowerBook G4 (Al, RevA, 1.25GHz, etc) didn't feel snappy and it annoyed me. While my friend's TiBook 1GHz that always felt snappier (depsite my faster CPU, GPU, bus, etc), CPU benchmarks always said my machine was slightly faster (as expected). However, up until now we had different configurations, different OS versions, different RAM, etc. Now that our machines are running same OS, configured the same way, with the same amount of RAM, I had to look into this.

    I've figured out that the system is much slower when operating on Processor: Reduced. While I knew this reduced CPU speed by at least 1/3rd, what surprised me is that this appears to reduce user interface performance much more. By a factor of 3-5.

    Freshly booted and processor set to Highest, XBench reports UI scores of about 170-190 on my system. That number goes down to about 100-140 if the machine has been running for a while and there are various background apps/dashboard widgets running. However, the machine still feels snappy to me even when benchmarking at 100.

    When I drop down to Reduced, my XBench UI numbers range from 20-75. In typical situations where other apps are running in the background, XBench UI numbers can drop to almost 20 (compared to the 100 as above). I am not directly bothered that the numbers are low. Rather the low numbers quantify my subjective feelings about the system. In Reduced, everything seems to lag and annoying sub-second pauses are frequent.

    I have been operating at Reduced because this thing gets so hot, and my recollection from testing with simple Apple tools is that Reduced only slows it down by 1/3rd (taking it from 1.25GHz to 800 or 867 or something like that). More recent tests with XBench, however, suggest that CPU speed is cut in half. Even then, not all system components are cut to half speed and the CPU isn't the bottleneck for everything the system does. One would expect that most things slow down less than 50%.

    And XBench numbers jive with that theory for most things -- while the GCD Loops performance drops almost exactly 50% going from Highest to Reduced, System Fill performance only dropped about 30%. Most of the other tests show drops of 35-45% with Reduced. UI always drops much more with Reduced (60-80%).
     
  2. stridey macrumors 65816

    stridey

    Joined:
    Jan 21, 2005
    Location:
    Massachusetts, Connecticut
    #2
    Yes, of course. That's the whole point of Reduced. Less power (ie slower), but less power (ie better energy usage/longer battery life). :)
     
  3. bzgnyc thread starter macrumors newbie

    Joined:
    Oct 19, 2005
    #3
    Of course -- but my point is that Reduced is much slower. The system should be 30-50% slower. The UI seems to run 60-80% slower.
     
  4. stridey macrumors 65816

    stridey

    Joined:
    Jan 21, 2005
    Location:
    Massachusetts, Connecticut
    #4
    Ah. Well, I'd say it seems around 30-50% slower to me, but it's hard to measure these things, you know? Plus, I've never really given it all that much thought. :p :)
     
  5. bzgnyc thread starter macrumors newbie

    Joined:
    Oct 19, 2005
    #5
    Actually, not so hard. XBench 1.2 (and perhaps earlier) includes a tool to measure such things. In the report from XBench, there is a line "Elements" under the section "User Interface Test". All the numbers listed in sections 2 and 3 of my post come from XBench. Perhaps if you read those sections, the point of my post will make more sense.
     
  6. ITASOR macrumors 601

    ITASOR

    Joined:
    Mar 20, 2005
    Location:
    Oneida, NY
    #6
    To tell you the truth, on my iBook, nope, never noticed a difference. I have it set to go on reduced when I'm on battery and it doesn't even come to my mind that it feels slower or anything. Even if I play a few of my games, I get the same or close FPS.
     
  7. mad jew Moderator emeritus

    mad jew

    Joined:
    Apr 3, 2004
    Location:
    Adelaide, Australia
    #7
    Yeah, I don't notice any difference in GUI performance on my iBook either, no matter which setting I choose. The speed difference is noticeable to me in more intensive tasks only. :)
     
  8. HiRez macrumors 603

    HiRez

    Joined:
    Jan 6, 2004
    Location:
    Western US
    #8
    XBench is crap. I would not pay attention to anything it says in regards to meaningful real-world performance. Read almost any Mac review. They always throw in an XBench run and then immediately explain how useless that test is.
     
  9. bzgnyc thread starter macrumors newbie

    Joined:
    Oct 19, 2005
    #9
    I've heard that and that is why I more or less ignore all its other results. However, one can watch the system benchmark the UI and OpenGL subsystems and the difference between Highest and Reduced is dramatic (at least on my system). Also, note that I used XBench to confirm and quantify my experience -- not the other way around.
     

Share This Page