New Power Systems Are Not Based on Power6+ Chips
October 13, 2008 Timothy Prickett Morgan
If you were paging through the Power Systems section of the IBM Website last week, you might have been convinced that the new Power 560 midrange server was based on the Power6+ processor, not the Power6 chip.
This screen certainly confused me, and a whole bunch of readers who emailed me last week:
As you can see, it clearly says Power6+, not Power6. I checked with my sources at IBM, and the Power 560 is not based on the Power6+, but rather the existing Power6 chip, which is implemented in 65 nanometer processes. And IBM is not, as it did with some pSeries and System p machinery in past years, cramming two Power6 chips (for a total of four cores) into a single processor socket, side by side. All of the updated Power Systems machines have processor cards or motherboards that have twice as many sockets on them, thus yielding the doubling up of cores in the box.
There is a Power6+ chip coming, by the way, which I told you about back in July. My gut says that with Power6+, IBM will cram two Power6 chips into one socket and pretest the 45 nanometer processes that it is perfecting for Power7, which is due in early 2010. The roadmap in that story clearly says that Power6+ is a multicore chip instead of a dual-core chip, and that it will offer roughly twice the oomph of a Power6 chip. The chip is also expected to run at a fast clock speed, which also suggests a process jump, too.
You’ll just have to wait a while longer for that. Hopefully Power6+ chips will plug into Power6 systems, or processor cards will be swappable between the two generations, allowing customers to upgrade their processing capacity without having to buy whole new machines. This is what is really important to most midrange shops.