First Look: A maximum look at a mini Mac, part three

EDITOR’S NOTE : The following is a series documenting Macworld Senior Editor Rob Griffiths’ first week with an Intel Mac mini. You can view each individual series installment:

  • Setup, configuration and application tests
  • General observations, audio & video, gaming
  • Testing methods, Intel transition and conclusions
  • More RAM, more tests
  • HD issues and final thoughts
  • VII. Testing the Mac mini

    You can see the results of the Macworld Lab’s tests in our full review. Since my personal objective was to learn as much about the Mac mini as possible, I sought out additional test tools to add to what the Macworld Lab tests told me.

    Geekbench benchmark: Impressed with the relaunch speeds of applications, I went looking for a benchmark test that would measure the raw computing power of the Core Duo chip in the mini. With some help from Google, I found Geekbench, which runs in the Terminal and runs a whole slew of number-crunching tests (and works on Windows and Linux, too). I ran Geekbench on all three Macs. One interesting feature of Geekbench is that there’s a Rosetta version as well as a native version, so you can see just what kind of performance impact Rosetta makes.

    Geekbench is, well, quite geeky, so dig into the table below only if you want to see all the gory details.

    Rob’s Geekbench Results

    Intel Mac mini (rosetta) Intel Mac mini (Native) Rosetta as % of Native PowerBook Dual G5 Native as % of Dual G5
    Emulate 6502
    (1 thread)
    71 152 47% 65 154 99%
    Emulate 6502
    (4 threads)
    136 290 47% 70 282 103%
    Blowfish
    (1 thread)
    (see note) 291 N/A 184 143 203%
    Blowfish
    (4 threads)
    (see note) 494 N/A 181 253 195%
    bzip2 Compress
    (1 thread)
    49 115 43% 79 132 87%
    bzip2 Compress
    (4 threads)
    101 219 46% 76 240 91%
    bzip2 Decompress
    (1 thread)
    57 115 50% 66 141 82%
    bzip2 Decompress
    (4 threads)
    117 215 54% 65 269 80%
    Mandelbrot
    (1 thread)
    62 123 50% 64 131 94%
    Mandelbrot
    (4 threads)
    118 234 50% 62 247 95%
    Latency 210 432 49% 73 222 195%
    Read Sequential 152 306 50% 38 213 144%
    Stdlib Allocate
    (1 thread)
    (see note) 119 N/A 156 114 104%
    Stdlib Allocate
    (4 threads)
    (see note) 139 N/A 157 117 119%
    Stdlib Write 134 135 99% 34 132 102%
    Stdlib Copy 156 158 99% 45 134 118%
    Stream Copy 79 108 73% 34 135 80%
    Stream Scale 62 109 57% 33 133 82%
    Stream Add 82 162 51% 18 138 117%
    Stream Triad 70 158 44% 18 135 117%

    Noted results differed significantly from native tests, suggesting a possible bug in the test program.

    Testing by Rob Griffiths using GeekBench .

    If you just want the summary version of the results, here it is:

  • In nearly all tests, there’s about a 50 percent hit in performance due to Rosetta when compared to running in native mode.
  • My PowerBook is in a different (much lower) league when compared to the mini. In only a couple of tests did it score better, and in most of the others, it was resoundingly trounced.
  • In just over half the tests, the Core Duo chip was actually quicker than my Dual G5, and in the other tests, it was never far behind. I find that fairly impressive.
  • Clearly the Core Duo is a powerful chip, based on its raw number-crunching abilities. As more applications go Universal, we (consumers) will see the benefits in terms of improved performance.

    Cinebench benchmark: The free Cinebench benchmark uses the Cinema 4D engine to test the graphics performance of your Mac. I tested all three of the Macs here, and then, for an added data point, also tested my homebuilt Windows XP PC (which literally hadn’t been powered up in months). Here’s how the machines performed:

    The first thing that stands out, quite glaringly, is that the Windows XP box kicked some serious Cinebench butt on the OpenGL benchmarks—it was over twice as fast at the hardware accelerated test! Now Maxon’s OpenGL engine implementation may not be the best, but the fact is that Cinema 4D’s OpenGL engine will run twice as quickly on my homebuilt Athlon-based single-core CPU as it will on my Dual G5. That just doesn’t seem right.

    Also obvious from the charts is that the 12-inch PowerBook is really not a great machine for Cinema 4D work. It was substantially behind the mini on all tests.

    What do these results mean to you, if you don’t work in Cinema 4D all day? Not necessarily a whole lot, other than to put the relative performance of the machines in perspective, and to note that the mini’s onboard graphics chip works much better than the PowerBook’s separate video card.

    Xbench: This is one of the older Mac benchmarking applications, and it too is now available in Universal form. I ran all three machines through the standard test, and here are the results:

    Rob’s Xbench Tests

    Intel Core Duo 1.66GHz PowerBook G4 1.33GHz Power Mac G5 2GHz
    CPU 63.94 55.59 100.14
    Thread 164.8 58.38 101.17
    Memory 94.66 25.91 91.58
    Quartz Graphics 57.99 62.17 101.16
    OpenGL Graphics 184.9 68.73 116.16
    User Interface 23.83 34.23 90.17
    Disk 30.25 26.38 50.86
    Overall Score 53.87 40.83 87.46

    Testing by Rob Griffiths.

    Xbench is designed to return a score of 100 on a 2.0GHz Dual G5, which just happens to be my desktop machine. As you can see, my system did better in some areas like graphics (where my XT800 card is quicker than the stock card), but much poorer in others—I completely failed the disk test, for example. I’ll have to look into that at some point!

    The interesting figures here are the three tests where the mini beats the Dual G5. The Thread and Memory results don’t surprise me, given the much faster RAM and the Intel chip’s capabilities. The OpenGL results, though, I have no explanation for. My Dual G5 is clearly much faster than the mini at anything using OpenGL, but these results suggest otherwise. And yet, in something like the iTunes visualizer, running at the same resolution on both machines, the Dual G5 is about 50 percent quicker than the mini. I really don’t have an explanation for this result, and welcome any thoughts from others on the subject.

    Three takeaway points:

  • Rosetta has about a 50 percent raw performance hit, though its actual impact will vary by application.
  • In many raw CPU performance tests, the Intel Core Duo chip outperformed the 2.0GHz Dual G5s in my desktop Mac. Not bad at all, given that this Core Duo mini is about one-fourth the cost of a high-end G5.
  • The PowerBook G4/1.33GHz is completely dominated by the Core Duo mini in these benchmark tests.
  • These test results are more for general interest than specific comparison purposes, especially as they lack any comparative info for the prior-generation mini. Still, they do show that the mini has a powerful CPU and not-too-shabby graphics chip, both of which should help it easily fulfill its role as Apple’s entry-level Mac.

    Subscribe to the Apple @ Work Newsletter

    Comments