I am somewhat dubious about
VirtualPC. More than anything, I see it as a last resort: something to fall back on if there is absolutely no tool available to run natively under OSX to get the job done. Getting excited about VPC7 really suggests that, performance-wise, getting an Apple computer may have not been a very rational choice.
Even if graphics are offloaded natively to the GPU (speaking of which - which GPUs will be supported? Nvidia, ATI, all of them...?) there will still be an atrocious performance hit when compared to running the same OS/application/game combination on a similarly priced PC. I expect that playing any PC games less than a year old will be a lacklustre experience even on top-grade G5 hardware.
Also please remember that running an application under VPC requires you run the OS too - this may be stating the obvious but it is a fact that is often overlooked. Personally, I find the
Darwine Project to be far more promising, because it does not require that you run the OS in the background. It simply translates API calls. This does not necessarily mean that you don't have to possess a valid Windows licence (it still requires some Microsoft Windows .dll files that are subject to the EULA) but at least you don't need to run the whole kernel, all the Windows services, network stack, etc. on the background on your machine just to run, say, Microsoft Project. Darwine may not offer the lowlevel optimisations Microsoft is apparently implementing (at least not yet) but, native graphics aside, it will be a far more lithe system for running
x86 applications on the Mac. It's also free, which is always an advantage.
I do of course keep a copy of VirtualPC 6 on my machine, and I expect I'll find myself buying a copy of VirtualPC 7 shortly after it is released - mainly as a fallback solution. But as an occasional
FreeBSD user, I have grown to appreciate the flexibility of WINE and so that's the solution I'm really looking forward to.