thatwendigo, why are you trying to compare the power usage and thermal output of the Power Mac G5 with my hypothetical, 970fx-based iMac? I'm using the only available 970fx machine (the new xServe G5) as an example of what might be possible given the new, lower-power 970fx.
I think it is pretty obvious that given the power consumption and thermal output of the minimum configuration G5 xServe that a 970fx-based iMac is at least a possibility (particularly if it runs at a clock speed below 2GHz). However, I think we all agree that you couldn't put a Power Mac G5 into an iMac case (but now you seem to want to use that as the standard of comparison rather than a 970fx-based system). Since the Power Mac G5 uses the older PPC970 processor while the G5 xServe uses the 970fx couldn't that be the reason why the xServe design uses so much less power than the Power Mac? And as I've already said, adding a PCI graphics card to the xServe only adds 10 watts. In any case, it's likely that an iMac would use a relatively low-power graphics chip set -- perhaps even one designed for mobile applications.
And I have to disagree with your claim that the Power Mac G5 and Athlon 64 are actually closer in FSB design than is the Power Mac to the Pentium 4. If you disregard the 64/32 bit issue (which I think is irrelevant for this part of the discussion), then the only significant similarity between the Power Mac and the Athlon 64 system designs are that both implement a bi-directional HyperTransport link as part of what could be termed the system bus. However, the HyperTransport link on Athlon 64 is connected directly to the processor, while the G5 connects to the HyperTransport bus through the Power Mac's system controller (or the so-called northbridge). In fact, I thought that we had already agreed that the HyperTransport link on the Power Mac was not a part of the processor's FSB.
Look at it this way. On the Athlon 64 the FSB consists of a direct link to system memory and a direct connection to the HyperTransport bus. However, on the Power Mac G5 neither the memory nor the HyperTransport bus are directly connected to the processor. On the Power Mac the FSB connects to the system controller which in turn is connected to the memory and HyperTransport bus (i.e. the system controller sits between the processor FSB and the memory and the HyperTransport bus).
However, since both the Power Mac and Pentium designs use the FSB to attach the processor to the system controller (northbridge) I think the FSB on these two processors is really quite similar in purpose. In fact, please take a look at the diagram at the following link:
http://www.aliusa.com/eng/product/corelogic/m1681.htm
Here you will see a Pentium 4 chip-set implementation that looks almost exactly like the single-processor Power Mac G5 design (including a HyperTransport bus running off of the system controller).
It seems that those of you who are arguing that it should be relatively easy for the G5 to support FSB speeds similar to the Athlon 64 and Opteron are overlooking Apple's custom-designed system controller. The system controller is not some passive device that just sits between the processor and the remainder of the system. To reach the FSB speeds you are suggesting you need a processor, system controller, and motherboard design that work in very close harmony. I'm not trying to suggest that the system controller alone determines the limits for the FSB speed, but it can play a role and right now I think this is one of the unknowns about how fast the FSB can run on the Power Mac.
Thus, I'd say that the FSB potential on the current Power Mac G5 design is somewhat of an unknown. Could it run at 1.1GHz? Almost certainly. How about 1.2GHz? I'd say probably. And what about 1.3GHz? Who really knows? I expect that only Apple and IBM know for certain.
It would certainly be possible for Apple to redesign or refab their system controller so that the Power Mac FSB could run well in excess of 1GHz. However, for this next round of products I don't think we'll see a new system controller. That's one reason why I now think it is possible that we'll see a 3X bus multiplier in the new Power Mac G5s. Well, at least we may see that if the processor speeds exceed 2.4GHz (why 2.4GHz? Because a 3X multiplier fits so nicely if you need those speeds with a drop-in processor replacement).
I think it is pretty obvious that given the power consumption and thermal output of the minimum configuration G5 xServe that a 970fx-based iMac is at least a possibility (particularly if it runs at a clock speed below 2GHz). However, I think we all agree that you couldn't put a Power Mac G5 into an iMac case (but now you seem to want to use that as the standard of comparison rather than a 970fx-based system). Since the Power Mac G5 uses the older PPC970 processor while the G5 xServe uses the 970fx couldn't that be the reason why the xServe design uses so much less power than the Power Mac? And as I've already said, adding a PCI graphics card to the xServe only adds 10 watts. In any case, it's likely that an iMac would use a relatively low-power graphics chip set -- perhaps even one designed for mobile applications.
And I have to disagree with your claim that the Power Mac G5 and Athlon 64 are actually closer in FSB design than is the Power Mac to the Pentium 4. If you disregard the 64/32 bit issue (which I think is irrelevant for this part of the discussion), then the only significant similarity between the Power Mac and the Athlon 64 system designs are that both implement a bi-directional HyperTransport link as part of what could be termed the system bus. However, the HyperTransport link on Athlon 64 is connected directly to the processor, while the G5 connects to the HyperTransport bus through the Power Mac's system controller (or the so-called northbridge). In fact, I thought that we had already agreed that the HyperTransport link on the Power Mac was not a part of the processor's FSB.
Look at it this way. On the Athlon 64 the FSB consists of a direct link to system memory and a direct connection to the HyperTransport bus. However, on the Power Mac G5 neither the memory nor the HyperTransport bus are directly connected to the processor. On the Power Mac the FSB connects to the system controller which in turn is connected to the memory and HyperTransport bus (i.e. the system controller sits between the processor FSB and the memory and the HyperTransport bus).
However, since both the Power Mac and Pentium designs use the FSB to attach the processor to the system controller (northbridge) I think the FSB on these two processors is really quite similar in purpose. In fact, please take a look at the diagram at the following link:
http://www.aliusa.com/eng/product/corelogic/m1681.htm
Here you will see a Pentium 4 chip-set implementation that looks almost exactly like the single-processor Power Mac G5 design (including a HyperTransport bus running off of the system controller).
It seems that those of you who are arguing that it should be relatively easy for the G5 to support FSB speeds similar to the Athlon 64 and Opteron are overlooking Apple's custom-designed system controller. The system controller is not some passive device that just sits between the processor and the remainder of the system. To reach the FSB speeds you are suggesting you need a processor, system controller, and motherboard design that work in very close harmony. I'm not trying to suggest that the system controller alone determines the limits for the FSB speed, but it can play a role and right now I think this is one of the unknowns about how fast the FSB can run on the Power Mac.
Thus, I'd say that the FSB potential on the current Power Mac G5 design is somewhat of an unknown. Could it run at 1.1GHz? Almost certainly. How about 1.2GHz? I'd say probably. And what about 1.3GHz? Who really knows? I expect that only Apple and IBM know for certain.
It would certainly be possible for Apple to redesign or refab their system controller so that the Power Mac FSB could run well in excess of 1GHz. However, for this next round of products I don't think we'll see a new system controller. That's one reason why I now think it is possible that we'll see a 3X bus multiplier in the new Power Mac G5s. Well, at least we may see that if the processor speeds exceed 2.4GHz (why 2.4GHz? Because a 3X multiplier fits so nicely if you need those speeds with a drop-in processor replacement).
