Here are results of our testing using applications, as opposed to
'artificial' benchmarks. The 733 uses a new generation G4 with a longer
pipeline. Applications (and benchmarks) that have not been optimized or
compiled to support this pipeline may show different performance.
By the way, in the thread, several people erroniously confused disk cache
with processor cache, so said "turn down the disk cache to reduce cache
misses"--disk cache is optimized at the default setting, but only affects
reading files from the disk, not access to memory.
Photoshop 6.0, 10 operations
733MHz PowerMac G4..61 seconds
667MHz PowerMac G4..64.6 seconds
Dual 533 MHz PowerMac G4..64.7 seconds
533 MHz PowerMac G4 ..72.4 seconds
466MHz PowerMac G4..83.9 seconds
Media Cleaner 5.0, time to output movie using Sorenson codec
PowerMac G4 733..82 seconds
PowerMac G4 533 dual processor..58 seconds
PowerMac G4 466..103 seconds
Final Cut Pro, total time of several different operations
PowerMac G4 733..229.8 seconds
PowerMac G4 533 dual processor..173.3 seconds
PowerMac G4 466..309 seconds
Here is a reply by David Baker [on G4/733 testing]
It is worth noting the general thrust of the comments below is that many of
the tests in fact are limited by performance of other systems and are NOT
reflective of relative processor performance. Here is a list of issues
with the testing reported, based on things we have learnt went doing our
1) Quake 3 @ 1024x768 is primarily graphics card limited this is
not a fair test of CPU performance. If you want to see the effects
of CPU run at 640. [I had posted similar comments in the forums here last week, 640x480 mode helps rule out the graphics card as a bottleneck, although drivers can still affect Mac results of Radeon vs GeForce2MX.-Mike]
Try this with 533MP (OS 9 and OS X) and 733 (OS 9), you will see significant performance improvement between the 533MP (OS9) and 733(OS9). Obviously when you use the OS X with support for MP you will see another improvement. I should note in our PC testing we can see variances of up to 20% due to different drivers in the same machine..
2) Unreal, as I stated above 1024 tends to be a graphics card limit [not]
a CPU limit. Since the Mac's are consistently 50% the performance of
the PC's one would expect that the issue here is RAVE vs. DirectX.
3) iTunes. As we have found out that any MP3 rip from a CD is
dependant on the spindle speed of the CD-drive. [This is why in past reviews I've noted Soundjam tests from the CD and the hard drive.-Mike]
If you wish
to measure CPU speed try ripping an AIFF file off your hard
disk. This is why I am not surprised that the 533 is equivalent
to the 533MP. I imagine that the Superdrive in the 733 is running
at a slightly slower speed for CD-Audio read than a CD/DVD-ROM.
Be aware that Mac's dynamically adjust the CD speed to allow for
the balance of the CD, given very small factors can impact this
you can see variance in test results so depending on the drive
in the machine, dirt on the CD and a whole host of issues there
will be variance between drives and different machines.
4) Photoshop, depending on what you test you results can vary.
If you solely test filters you are testing something like 5
different code paths. Lots of the filters are based on the
same core logic with different parameters. If you want to see
the effect of CPU speed try CMYK->RGB conversion we found it
scales with CPU and is NOT MP aware. So what we test is filter
performance AND many of commonly used functions.
Anyway let me know if any of this is helpful and whether there is
anything else you would like to know...
Glenn (& David)"
A G4/733 owner with Radeon card sent his PSBench test results (w/PShop 5.5) and the RGB->CYMK filter times were significantly lower than a single G4/500 (almost half) and faster than a 1GHz PIII for that conversion. (The total time for all 21 filter tests were also faster than either of those systems.) However many G4/733 owner's have reported lower than expected performance in real applications, not just benchmarks, using an array of Photoshop filter tests, iMovie tests, etc.
Update: For more detailed performance tests of the G4/733 with a GeForce3, as well as comparisons to a dual G4/500 see the most recent G4/733 and GeForce3 review. Includes comparisons of Radeon vs GeForce3 graphics cards in both systems also.
To see how a G4/733 w/Radeon AGP card fared against a 1GHz Pentium III, a single and dual G4/500 and a Beige G3 with dual G4/500 cpu upgrade in Photoshop 5.5 filter tests, click here.
7450 G4 CPU Chip Tech Docs and Errata: I've always wondered if the errata (errors) in the initial 7450 chip versions could be a factor in its performance. It appears this is so from the errata noted in some Motorola 7450 chip documents. See this 3/19/2001 news story for links and details.
(Update - the later 2.1 rev of the 7450 fixes some errata, but there's mixed reports on improved memory bandwidth, a sore point based on benchmark tests of the 7450 vs 7400/7410s.)
Several G4/733 owners have posted their apps tests results in this (long) forum thread which also has comments on compilers, the 7450's longer pipeline effects on performance and more. You have to read all the posts or at least scan every one of the pages in that thread to get the full picture. (Not just benchmarks are noted, but actual apps tests.)