> Point one to the G5, a few more PCI-E lanes than the X1000

Yes, 32 vs. 18.

> Do you really think emulated floating point code is going to slow a 1200MHz computer
> down below the performance of a 400MHz computer?

We know that floating point performance of code compiled natively for the SPE of the 1.2 GHz e500v2 is about that of a 600 MHz G3. Depending on the efficiency of the FPU emulation, I think it could end up undercutting a 400 MHz G2. But we'll see, I guess.

> I suppose it would depend on the balance of the instructions in the programs they
> were running

Yes, absolutely. My rather pessimistic estimation is about pure floating point code, of course.

> Which do you think would be faster, an Efika or that 366MHz 604e power A4000?

http://morph.zone/modules/newbb_plus/viewtopic.php?forum=11&topic_id=11137&start=744

> a 604e is going to be faster than the 603e related processor in the Efika.

Indeed, but the Efika is running at 9% higher clock and isn't crippled by a slow memory interface (which the PPC G2 CPUs are on the PowerUP boards in order to accommodate to the m68k CPU's needs).

> but a 366MHz A4000? Why would you ever sell that?

Well, I sold my 200 MHz A4000 when I found myself not using it anymore after having bought a 600 MHz Pegasos I. I never regretted it :-)

> Running OS3.1-3.9 it would actually be fast enough to compete with our slower offerings.

I doubt it. Being an m68k-based OS, OS3.x uses the PPC CPU only as kind of co-processor through specially written software. Beside the crippled memory interface, there's a rather big speed penalty caused by context switches (cache flushes) whenever a PPC program goes through the m68k OS API.