[PEHowland]

Hello all,

I shall shortly be writing a review for Sudhian Media in which I compare the performance of the iMac 1.8GHz with stock 512MB memory versus the same iMac with dual 512MB memory from a leading memory manufacturer. The specifications of the 3rd party memory is better than Apple's own, and of course the machine will be running in dual-channel mode.

I have benchmarked many PC's over the years, but this will be my first review of this type on an Apple.

So, my question to you all is what benchmarks or applications or aspects of usuability would you like me to assess within the review? What's the best way of assessing memory performance on an Apple? What would you like to see tested?

Many thanks in advance for this. I can of course figure out some tests on my own, but I thought it would be more interesting to get feedback from the Apple community before I start to make the review as relevant as possible.

[turtle777]

Sorry, can't help you on that.
But please keep us posted about the results.

-t

[PEHowland]

Originally posted by turtle777:
Sorry, can't help you on that.
But please keep us posted about the results.

Thanks, I will do. The review won't be for a couple of weeks yet - the iMac is not due to arrive until 4 October.

So far, some benchmarks I'm considering include:

• PSBench for Photoshop performance
• Mathematica benchmarks for scientific work
• Linpack Java Benchmark for Java performance
• Java array test
• SciMark 2.0 for more scientific performance tests
• XBench
• STREAM benchmark for measurement of memory bandwidth
• Some kind of MP3 ripping test using iTunes
• CocoaBench to measure Cocoa performance
• CineBench 2003 G5 beta for 3D rendering performance [probably more GPU influenced than memory, but maybe of interest]
• Time to boot
• Quake III frame rate test (at low resolution to put emphasis on memory/CPU subsystem rather than graphics)

Any other ideas for assessing the impact of more memory and also faster memory would be much appreciated.

Thanks,

[klinux]

Does Apple hardware test CD test speed of memory?

[Pierre B.]

Originally posted by PEHowland:

Any other ideas for assessing the impact of more memory and also faster memory would be much appreciated.

Not exactly what you asking for, but you could consider AltiVec Fractal Carbon, running with Altivec code on and off. Don't forget to report values of colour speed and maximum count (displayed in the corresponding menus). Increasing the maximum count (default is 4096), results in more complicated calculations that take much more time to complete. You can try values more than 100000; this should be enough to put some serious stress on the system. It would be interesting to see how the results (displayed in Mflops in the AltiVec Fractal Carbon window bar) are affected by different memory configurations.