UPDATED 1/3/05 with 800 G4 iMac data.
UPDATED 11/28/04 with 1.8 GHz G5 iMac (20") performance data.
UPDATED 1/24/04 with AMD 2800+ CPU performance data.
UPDATED 10/25/2003 with Dual G5 2.0 GHz and 1.8 GHz Cinebench data using an optimized Cinebench (beta).
It doesn't matter whether you've been reading the Mac or PC magazines. You probably know that some of the big news has been Apple's release of their up and coming G5 (IBM 970) powered PowerMacs. I got my hands for a few minutes on a dual 2.0 Ghz PowerMac, a 1.6 GHz PowerMac, a 1.8 Ghz G5 PowerMac (rev A), and a 1.8 GHz G5 iMac (20"). I was curious to see how the G5 would stack up in and against my personally owned crop of machines which includes a homebuilt AMD powered 2800+ CPU running Windows XP. Using Cinebench 2003, I felt it was an execellent opportunity for me (and you!) to see how it stacked up against both other Macs and a Windows machine if I bought one and brought it home from the store. So, I downloaded Cinebench, ran the benchmark, and tallied the results up using Microsoft Excel.
(NOTE: The version of Cinebench I'm using as of 11/25/04 is optimized for the G5 but is still called a beta release. When I first posted this article, rumor had it that more work by Maxim on the benchmark might push up the results even higher. We'll have to see if they release a new version before Apple hits the streets with a true 64 bit operating system. Personally, I think we're not going to see any differences in the benchmark until that happens.
Since I was hijacking showroom G5's for the PowerMac portion this test, this test set for the 1.8 and dual 2 GHz run with the Energy Saver setting in Automatic. The Energy Saver settings were locked where I could not change them. [If you're wondering why I'm talking about that, see The Computer Blog entry about G5 bus slewing.] Using the "Highest" setting will yield slightly better results. For the 1.8 G5 iMac test, Highest was used. Keep in mind, though, that the 1.8 PowerMac's bus speed is 900 MHz and the 1.8 iMac's speed is 600 MHz. That's probably the biggest reason for the small gains that the 1.8 Ghz PowerMac shows over the iMac in these charts.
I've left the old, un-optimized Cinebench code benchmarks in these charts as a reference. The optimized code shows performance increases laid in the 20% - 26% range. Depending upon the variable examined, actual performance increases ranged from as little as 3% to as high as 31%. In some cases, there was a performance loss. I believe this was most probably due to some factor in the benchmark code of the optimized version (though it also might have been because of CPU throttle back due to Energy Saver settings). For instance, in the Open GL tests, scene 1 test results were always slower than previous results in the same scene while scene 2 results were always higher than previous results in the same scene.
One benchmark is never the full story. See the G5 Benchmarking page to see what other benchmarks are showing. And stay tuned. The G5 will be an evolving story...
Here are the configurations of the computers I used in the test, other than the G5: (1) Apple Quicksilver 2001dual 1 GHz CPU PowerMac, 1.5 GB PC133 SDRAM, two 120GB HD's (7200 and 5400 rpm), Radeon 9000 Pro video card; (2) Apple 2003 Mirror Door dual 1.25 GHz PowerMac, 1.2 GB DDR RAM, 160 HD, 120 HD (7200 rpm), Radeon 9000 Pro video card; and (3) homebuilt AMD 2000+ (1.67 GHz) CPU, 512MB DDR RAM (PC 2100), 60 GB HD (5400 rpm), 120 GB HD (7200 rpm), and ATI All-In-Wonder 9000 video card, and (4) the same AMD system upgraded to an AMD 2800+ CPU. The Macs were running OS 10.2.6. The AMD powered machine was running Windows XP Home. The 1.6 GHz PowerMac G5 was stock except it had 768MB of RAM and used Jaguar (OS 10.2). The 1.8 Ghz PowerMac G5 had 512 MB of RAM in both optimized and non-optimized test sets. The Dual 2GHz PowerMac G5 machine had 512MB RAM in the optimized test set and 1 GB RAM in the non-optimized test set. The dual 2.0 GHz and 1.8 GHz G5 benchmarks were run with the Energy Saver CPU setting on "Highest" in the non-optimized tests and"Automatic" in the optimized tests. (NOTE: I re-ran the G4 PowerMac Cinebench tests using Panther. The differences laid in the 1%-2% range, so I considered them insignificant factor and did not update the G4 results.) The 1.8 GHz G5 iMac had 1 GB RAM with a Nvidia GeForce FX 5200 GPU with 64MB video RAM) and was run with Eneregy Saver at "Highest" under Panther (OS 10.3.6). The 800 G4 iMac had 512 MB RAM and used Jaguar (10.2.8) as an operating system but otherwise was factory "stock".
The posted results include the Open GL hardware and software lighting tests, the cinema 4D shading tests, CPU rendering tests, and a compilation of the final Cinebench benchmarks. Cinebench 2003 was used for these tests. (NOTE: The Cinema4D Shading Polygon tests include only the G5 1.6 GHz's data due to an oversight on my part when recording the data. An "Opt" designation beside the machine name means the test was run using Cinebench 2003 beta optimized for the G5.)
Above: The 1.8GHz G5 with a 20 inch Apple Cinema Display.