Are there any comparisons between the various PPCs? I'd be curious to see how the PA6t, G4 and Titan cpus stood.
I'm sure google will bring up something.
-edit-
Off the top of my head, the last "G4" processors used in apple machines were the 7448 Apollo 8 variant. Now I don't think Apple ever used them directly as these are Freescale parts but 3rd party vendors have released upgrades for G4 boxes using them. A quick search reveals that these:
Are fabbed at 90nm process, copper interconnect
Single core, 32-bit
Have altivec
Run up to 2GHz with 200MHz FSB
TDP of ~18W @ 1.7GHz
1MB L2 cache
The PA6T:
Are fabbed at 65nm process
Dual core, 32/64-bit
Have altivec
Run beyond 2GHz (haven't found a direct source for the ceiling, but 2.5GHz is quoted a lot).
TDP of ~7W @ 2GHz
Upto 8MB L2 cache (configurations from 512K upwards) with "MOESI" cache coherency (see AMD64 docs for more on that).
Virtualisation/Hypervisor features
Dual integrated DDR2 memory controllers at ~1066MHz
What was interesting about the X1000 motherboard was the presence of 2 pairs of DDR slots, the arrangement of which would seem to lend itself to the dual DDR2 configuration of the above PA6T specs.
One thing that doesn't stack up is the instruction set architecture version. AFAIK, the PA6T adheres to the v2.04, but the documented info on the X1000 suggests a v2.05 conforming processor is used. It could be there is a newer revision of the PA6T kicking around.
It might explain some of the cloak and dagger antics over the CPU.
-/edit-
Sounds almost like running a VM.
Well, not really, that's a level beyond even. Many SMT capable OS allow a process to set an affinity for a given core, meaning that that process will run on that core and will never be rescheduled to run on another during it's lifetime.
Such behaviour may be one way of keeping legacy processes together. Anyway, that's pure speculation on my part as I was having a think about how I might tackle the problem.