5GHz is screaming fast as it is, many will say clock cycles do not really matter. However, the fact-of-the-matter is that they do within a particular architecture. And we already know the Power architecture is very efficient. What the Power6 brings to the board is a top end 2x plus frequency boost over the popular Power5 series. This is likely to increase with process improvements. Plus many architectural changes that both increase efficiency and lower power consumption.
Don't expect this chip in your next Mac. It targets the server market, where it should do very well, due to its vastly increased performance over the Power5 and lower power consumption architecture, a large if not primary factor in data center operation cost.
"The Austin-engineered chip will be massively complex, with 750 million microscopic transistors. And it will be race-car fast, running at about 5 gigahertz, about twice as fast as competing server chips. Analysts say that IBM is well ahead in the race for high-performance server technology and that the Power6 chip will simply increase its lead. Its predecessor, Power5, ran at about 2 gigahertz when it was launched in 2004. (The speed of server processors generally lags those of PC chips because they are designed and used differently."
I haven't been able to get the guys down the road to invite me to an extreme Frisbee match yet, or a beer blast, but these guys have produced a true next generation processor. Currently on a 65nm process, I would expect a feature shrink to 45nm sometime over the next two years, which will both decrease power requirement and increase performance and yield.
The IBM Power chip series has be criticized for its power usage and heat dissipation requirements in the past; this chip promises to reverse that trend. It appears as though the Power6 will quickly reinstate IBM as the front runner in large scale server operations. The virtualization capabilities of the chip are yet to be independently benchmarked in real-world application and the yield question is un-addressed at this time. Will IBM be able to manufacture enough of these at a sufficient yield to meet demand? This will likely spell the success or failure of the Power6 -- time will tell.
