Slashdot Mirror
Apple Updates Power Mac Line
Phreck writes "Apple has announced an upgrade to its Power Mac line today. The new Power Macs all feature dual G5 processors, 512 MB RAM, and dual-layer 16x SuperDrives. On the low end is the dual 2.0GHz with 160GB HD and ATI Radeon 9600. The mid-range includes dual 2.3GHz processors with 250GB HD and ATI Radeon 9600. The top-end system has dual 2.7GHz processors with 250GB HD and ATI Radeon 9650. The processors are not the dual-core variety as has been rumored for weeks now."
Perhaps you should start visiting the rumor sites. this powermac update was not a surprise
MacRumors.com
MacRumors' Buyer's Guide also keeps track of time since last release and a summary of recent rumors, and a buy or not-buy recommendation.
Put simply, dual core means that both CPUs are on the same piece of silicon. They can share a unified cache, access it faster, and resolve deadlocks & invalidates etc much faster.
A dual core processor will also run cooler than two single cores, and the reduced number of external interconnects means that the whole thing can be clocked faster.
Since you are using up to twice the wafer size, you need to have a high yield rate of you're going to keep costs down: Yield decreases in proportion to wafer area.
It's worth reading up on System On Chip design - see how you can put the graphics controller, DSP, and USB controller on the same wafer. Furber's ARM SoC book is slightly dated but nevertheless a good read.
Relative to the latest AMD etc depends on the code you're running. PowerPC has a lot of registers, can do much more complicated floating point arithmetic, and has a fused multiply-add instruction (good for FFTs) but in pure integer throughput the latest AMD etc will probably triumph.
Dual core being "more efficient" depends very much on the task being considered.
For any sufficiently large task, the bottleneck is the path to main memory. For a given level of package & bus limitations, dual-core must use an amount of bandwidth to main memory to feed two processing units rather than one.
For tasks that fit in on-chip cache, of course, the bottleneck is processing, and dual-core can be a huge improvement, especially where the synchronization overhead would have to go off-chip in the case of dual processors, as you mention.