Something wrong with the new G4 ?
Here's what I found on the MacSlash forum. Any comments on this ? :
The calculation to get that 1GB/s from MaxBus is:
133 MHz * 64 bits / (8 bits /Byte) = 1064 MB/s
so now that they are overclocking the bus to 167 MHz, I would expect to see that MaxBus number to now be up to
167 MHz * 64 bits / (8 bits /Byte) = 1336 MHz
Furthermore, some interesting math from Steven DenBeste (a blogger):
1GHz / 133 MHz = 7.5x clock multiplier on the old dual 1gig
1.25GHz / 167 MHz = 7.5x clock multiplier is the same!
1GHz / 167 MHz = 6x clock multiplier on the new dual 1 gig
133MHz x 6 = -800 MHz-
Therefore, I would bet that Apple is merely overclocking these two fastest CPUs, while using the same 1GHz and 867 MHz chips they were before. Moto allows this, and the concept was something that PowerLogix was doing for a while. And if they ARE overclocking, that could have SOMETHING to do with why the performance isn't really any different...the "new" 1GHz is running the same core as the old 867 Mhz, just faster...the question is, are the cores identical except for clock? Probably not.
It would also mean that the two low end models are basically the same computer, with some switches flipped...anyone remember the 233/266 Beige G3 overclocking tricks? Ready for round 2? (I thought it was kinda funny that Apple would ship TWO different logic boards.) My buying advice, get a Dual 867 and wait for the hack...
The calculation to get that 1GB/s from MaxBus is:
133 MHz * 64 bits / (8 bits /Byte) = 1064 MB/s
so now that they are overclocking the bus to 167 MHz, I would expect to see that MaxBus number to now be up to
167 MHz * 64 bits / (8 bits /Byte) = 1336 MHz
Furthermore, some interesting math from Steven DenBeste (a blogger):
1GHz / 133 MHz = 7.5x clock multiplier on the old dual 1gig
1.25GHz / 167 MHz = 7.5x clock multiplier is the same!
1GHz / 167 MHz = 6x clock multiplier on the new dual 1 gig
133MHz x 6 = -800 MHz-
Therefore, I would bet that Apple is merely overclocking these two fastest CPUs, while using the same 1GHz and 867 MHz chips they were before. Moto allows this, and the concept was something that PowerLogix was doing for a while. And if they ARE overclocking, that could have SOMETHING to do with why the performance isn't really any different...the "new" 1GHz is running the same core as the old 867 Mhz, just faster...the question is, are the cores identical except for clock? Probably not.
It would also mean that the two low end models are basically the same computer, with some switches flipped...anyone remember the 233/266 Beige G3 overclocking tricks? Ready for round 2? (I thought it was kinda funny that Apple would ship TWO different logic boards.) My buying advice, get a Dual 867 and wait for the hack...
Comments
While barefeats is quite often a good source of data, I find I very rarely agree with their analysis. Doubly so in this case.
First off, the DDR model has only 1MB cache per CPU, while the QuickSilver has 2MB per. The fact that they still performed nearly identically in processor-bound tests indicates that the memory subsystem in the DDR model has actually been improved fairly substantially; if the system controller was of equal efficiency from the processor POV, the DDR model would have certainly tested worse than the QuickSilver.
The fact that everyone seems to be missing, however, is that the new high-bandwidth memory allows system components besides the CPU to perform DMA access without impeding bandwidth to the CPU. The ATA busses, AGP bus, PCI bus, and FireWire controller all feature DMA functionality, and this is the reason the improved path between the system controller and main memory is so important.
If you were to benchmark a machine running processor-intensive operations while additionally accessing ATA devices, the AGP card, PCI cards, and FireWire devices, you'd likely see the DDR machines running circles around the QuickSilvers.
And thus ends the rant.