Excel alternating use of cores under load

[jonbeans]

Version: 2008
Operating System: Mac OS X 10.5 (Leopard)
Processor: Intel

I have the iStat menus installed to monitor my MacBook Pro's temperature and CPU loads. CPU load is represented by two graphs, one for each core.

While waiting for a large workbook with lots (thousands) of references and math, I noticed that the processor load graphs mirrored each other. In other words, if the load on core 1 was increasing, the load on core 2 was decreasing. When Excel is the only major load, they're exact inversions of each other.

This seems like evidence of a bottleneck. I suspect Excel 2008 Mac is spending nearly as much time deciding which core to use as it is doing its actual job.

I'm posting this in the sincere hope that somehow this seemingly inefficient behavior is responsible for the comparatively slow performance, and that maybe something can be done about it.

(No, I don't believe my casual observations will be news to the QA team, but... worth mentioning I guess.)

 

[CyberTaz]

Your insights are definitely appreciated, but there's no "QA team" here to
see them -- this is a peer-to-peer newsgroup where none of us has anything
to do with the development of MS software.

Just a couple of points to consider, however:

The first being that Office 2008 was well on its way to the shelves when
Apple announced the switch to Intel processors. The second is that the app
doesn't make the decision on how to distribute its load - that's the job of
the OS, especially if the app isn't written for multi-core/multi-processor
systems. Third, once the code is written you can't just "switch it" nor can
the revision be made by an update.

Although future updates will most certainly strive to increase performance I
believe we'll need to wait for the next major release to see any significant
advantage taken of multi-core processor technology.

Regards |:>)
Bob Jones
[MVP] Office:Mac

 

[jonbeans]

Bob,

Thanks for the info and insight. Odd if the official message board is not monitored (i.e. read, not responded to), but ok.