Sorry I just realised that your post is under Logic 8 so that's probably what you use. I just followed the link from my email which goes directly to the message and I completely forgot about the new Logic 8/Logic 9 forums
Peter, what is your buffer set to? I suggest you set it to the maximum (largest) buffer size. If you do any live recording with Logic, investigate and learn how to do live monitoring from your interface instead of Logic to deal with the large buffer setting during live takes. This is the way I work all the time. All my Logic settings/buffers are as large as possible so I can get ridiculous amounts of DSP in my laptop. People usually run from this and seek the shortest audio buffer size, because configuration an audio interface (that can do live input monitoring) is a learning curve in an of itself. But you can wait on that for now and just test out your system with the large buffer to see if it helps.
I use an RME Fireface 400, which gives me the ability to monitor vocals and instruments live. In the audio driver window, I have monitoring unchecked. This means I can not monitor audio input through FX inserts in Logic, but I don't care. Once they are recorded, I can play the tracks back normally with FX. I have outboard distortion and reverb devices just so I can monitor with some rudimentary FX via the interface, knowing as soon as the dry tracks are recorded in Logic, I can get the exact mix I am looking for. Obviously, I'm not explaining the routing on the Fireface 400 in this response; it's a time-consuming subject. But try the large buffer setting at least (in the audio driver window) and see if it helps.
But that is not an issue even on my 5 year old G5.
I can find a number of situations where direct monitoring is not so great. One is where you punch in on a multitrack session where you have to have the exact same feel as the one you come from. It's almost impossible to judge this if the sound isn't exactly the same. Another one obviously is where you use ie the amp simulator or another soft-FX.
Peter Lemer has a 2.66 Mac Pro, he shouldn't experience this because of his Mac but it could be the soundcard of course.
Good point on the punch thing. Your guitar example is the reason I have an external Micro Cube for guitar. But honestly, I never really have a problem with punches, even with my setup. If worse came to worst, I could always disable the inserts on playback and match to that sound instead. I just can't justify the convenience of a short buffer for the loss of so much DSP. I don't know how people stomach it. I would need the latest tower to put up with it at all.
But it's not really a big deal since you rarely multitrack unless you're at the very start of a project so no plugins have been added yet.
Having said that I do a lot of pre-production that eventually moves into production meaning that a normal scenario could be a couple of guys coming here with a bunch of songs. We will then kind of rehearse the songs along with a drumbeat (everchanging until we get something that works for us). After having worked out the basic structures and grooves we bring in a drummer who plays to what we have demoed. At this point we do have something like 10-15 tracks and at least one instance of Addictive Drums and a couple of softsynths or EXS instances. So while running this we multitrack 12-16 tracks and it runs quite effortlessly on my 5 year old G5 with a 128 bufferrange.
Still in Peter Lemers case he has a Pro Mac and he mentions it happens even in small projects and that remains a mystery. He shouldn't have a problem running small projects with maybe even a 32 bufferrange. I completely agree with you that he would have even more power at higher settings but how much power does one need.
Of course, like I mentioned earlier, the soundcard is a big player in this game.
Peter, there is a link in the FAQ to an Apple document describing various conditions under which CPU overloads can occur, with some suggestions as to how to avoid them. It is FAQ 15, you can find it here