Mac Musings

Panther at the Low End, and What to Do with Jaguar

Daniel Knight - 2003.10.30

If you have two or more Mac users in the house, Panther can be a real bargain, ranging from $40 to $100 per license depending on the number of users. Thanks to a reader who paid the first $129, we bought the family pack and have installed OS X 10.3 on five Macs - my eMac 700, my wife's 14" iBook 600, a son's 12" PowerBook G4/867, and a pair of 333 MHz iMacs.

We all like it. Panther is definitely more responsive than Jaguar, which was an improvement over 10.1. Some reviewers are saying that 10.3 is responsive enough that Mac OS 9 users won't be able to complain about the speed.

Because of the number of systems upgraded, benchmarks were only run once. These are not the average of several tests, and because of this they may be a bit less precise than reports you'll find on other websites.

The eMac

I first installed Panther on the internal 40 GB drive on my eMac, since I don't otherwise use that drive. I figured that if there were any horrendous bugs, my main drive (an external 80 GB FireWire drive) would be unaffected. Sure, it's slower than my main drive, but this was just to make sure everything was working before I committed myself.

I only ran into one problem after the update - I can no longer log into gaming lounges at Yahoo! Games using Safari. The problem exists in both 1.1 and 1.0 (which was still on my main drive). I suspect it's the latest Java interpreter, and I've reported the bug to Apple. (Everything worked just fine under 10.2.8 and Safari 1.0.)

After a day or so of use, and after finding no serious problems reported on the Mac Web, I installed Panther on my external 80 GB FireWire drive on Sunday. This is a 7200 rpm drive with an 8 MB buffer, so it's a lot more responsive than the eMac's stock hard drive.

The only other software problem I've encountered is with System Optimizer X. The program mostly works, although you can no longer tweak Internet performance, but the scheduler is all messed up and tried to run SOX at every restart - until I removed it.

Benchmarks

I ran Let1kWindowsBloom and XBench 1.0.3 on my eMac before updating to Jaguar - and again after the update. I also benchmarked the external RAID array I use for backup, a pair of 80 GB 7200 rpm drives with 2 MB buffers in a single FireWire enclosure.

The eMac has 640 MB of RAM installed. Here are the test results:

Version          10.2.8   10.3.0
Let1kBloom       56 sec   37 sec  +66%
XBench
 CPU Test         81.10    82.85   +2.5%
 Thread Test      42.64    58.48  +37%
 Memory Test      51.63    63.49  +23%
 Quartz Test      73.58    87.09  +18%
 OpenGL Test      81.95    74.90   -8.6%
 User Interface   58.40    98.26  +68%
 Disk, internal   38.18    34.21  -10%
 Disk, external   70.86    69.94   -1.3%
 Disk, RAID       73.65    72.95   -1.0%

There are a few places where performance dropped, a couple where there's hardly any change, and five tests that show a huge improvement under Panther. Memory tests 23% faster, threading is improved by 37%, and graphics sees a big boost, except for OpenGL. Let1kWindowsBloom ran 66% faster, and the User Interface test in XBench reported almost the same improvement at 68%. Quartz graphics are 18% faster on this first generation eMac.

The faster external FireWire drives saw less of a performance hit under Panther than the slower internal drive. On the other hand, they're nearly twice as fast as the internal drive.

The iBook

My wife's 14" iBook is a bit memory starved at 256 MB, and I've suggested she get a 512 MB upgrade when she can afford it, which will bring the 'Book to 640 MB. The iBook has the stock Apple drive.

Version          10.2.6   10.3.0
Let1kBloom       93 sec   54 sec  +72%
XBench
 CPU Test         39.41    31.49  -20%
 Thread Test      44.74    53.67  +20%
 Memory Test      21.78    24.02  +10%
 Quartz Test      50.42    55.83  +11%
 OpenGL Test      78.65    77.84   -1.0%
 User Interface   47.61    71.48  +50%
 Disk Test        35.04    36.74   +4.9%

Again, the user interface is much faster than under 10.2.x - Let1kWindowsBloom benched a whopping 72% faster, and the XBench UI test saw a 50% improvement. As on the eMac, the thread, memory, and Quartz tests also show improved performance, although not to the same extent.

The surprises are a 20% drop on the CPU benchmark, which may be an XBench anomaly (XBench isn't terribly consistent, and a lot of that is due to the nature of the operating system itself), and an almost 5% improvement in hard drive performance.

The iMacs

It takes a long time to install Panther on a 333 MHz iMac - and with the stock 6 GB drive in one iMac, I had to remove all of the iApps to make room for the installation. At least 10.3 put all these apps back on the hard drive.

I only ran 10.2.8 tests on the iMac with the stock Maxtor 6 GB hard drive and 320 MB of RAM. The second iMac has a 7200 rpm drive and 192 MB of memory.

Version          10.2.8   10.3.0
Let1kBloom      115 sec   82 sec  +40%
XBench
 CPU Test         22.03    20.81   -5.5%
 Thread Test      24.49    29.95  +22%
 Memory Test      17.58    18.51   +5.3%
 Quartz Test      33.31    38.31  +15%
 OpenGL Test       8.72    10.60  +22%
 User Interface   19.05    50.87 +167%
 Disk Test        28.08    21.70  -23%

Although the iMac has an ancient video chipset (ATI Rage Pro Turbo) and only 6 MB of video memory, Panther makes the most of limited resources. The Quartz test runs 15% faster under Panther, OpenGL 22% faster, and Let1kWindowsBloom 40% faster. But the most impressive result of all comes from the User Interface test, where Panther is 167% faster than Jaguar at the tested 1024 x 768 resolution with millions of colors.

The CPU test is a bit slower under Jaguar, but memory is faster, and threading is improved by 22%. As noted with the eMac's internal hard drive, the slower the hard drive, the more it suffers under Panther. In this case, the Disk benchmark is 23% slower than under Jaguar.

Now on to the other iMac, which has a partitioned 20 GB 7200 rpm hard drive. In every test except for the Disk Test, the iMac with 320 MB of RAM outperformed the 192 MB machine:

Model            192 MB   320 MB
Let1kBloom      104 sec   82 sec  +26%
XBench
 CPU Test         14.94    20.81  +39%
 Thread Test      26.34    29.95  +11%
 Memory Test      15.31    18.51  +21%
 Quartz Test      31.83    38.31  +20%
 OpenGL Test       8.83    10.60  +20%
 User Interface   40.20    50.87  +27%
 Disk Test        38.76    21.70  -44%

As long as OS X has been available, the recommendation has been to install as much RAM as your Mac can handle. The iMac with the extra 128 MB of RAM won every comparison that didn't involve the hard drive, demonstrating the boosting from a little memory to a lot can easily provide a 20% or greater boost in performance. If you're running OS X with less than 256 MB, consider upgrading if your Mac supports it.

The only place where the second iMac shone was the disk test, where the Seagate Barracuda trounced the stock Maxtor drive in the first iMac. Even an iMac, with a 16.7 MB/sec. IDE bus, will benefit from a faster drive. The stock drive never hit 10 MB/sec. throughput on any benchmark, while the 7200 rpm drive benchmarked 79% higher, moving data over the IDE bus as fast as possible during some tests.

For faster booting, faster program launching, and faster saves, once you've boosted RAM, look into a faster hard drive. You'll definitely notice the difference.

The PowerBook

The fastest Mac in the family belongs to my son Brian, who started college this fall. He chose to buy a 12" PowerBook G4 as his school computer, and he tosses it into his backpack and bikes to school with it every day that he has classes. At 867 MHz, the PowerBook's CPU should benchmark about 24% faster than my 700 MHz eMac. (Although this was a current model a few months ago, it nearly qualifies as low-end, since Apple only makes one model with a slower clock speed.)

The PowerBook was not benchmarked before Brian installed Panther.

Version         10.3.0
Let1kBloom      29 sec
XBench
 CPU Test       104.13
 Thread Test     73.60
 Memory Test     82.36
 Quartz Test    105.27
 OpenGL Test     92.04
 User Interface 117.97
 Disk Test       54.53

Impressive performance. Too bad we didn't benchmark before the upgrade.

What About Jaguar?

Yes, we've all griped about the fact that Apple doesn't offer any discount for upgrading to Panther, except for those who bought G5s or picked up a new Mac in the past few weeks. The family pack helps if you have more than one Mac user in the house.

But there's one benefit to Apple's "no upgrades" policy - you may well have one or more full copies of OS X (10.1 or 10.2) that you can sell on eBay, put on an older Mac, or donate to a friend, a school, or a local nonprofit that uses Macs. Maybe we can add another million OS X users that way.