Well their system works the same way:QUOTE
Cray's implementation of the DCP architecture optimizes message-passing applications by directly linking processors to each other through a high performance interconnect fabric, eliminating shared memory contention and PCI bus bottlenecks.
What direct connect on a CPU means is that the I/O Bus links memory, cache, hypertransport and the CPU Core together.
If you run two or more CPUs they can "direct connect" to each other.
This means via hypertransport they can connecto to each other and use eachother's memory, extra hypertransport links, or even Cache without any extra chips or protocols.
In old dual or more CPU setups all the CPUs would connect to a "joiner chip" or a northbridge. They would share memory, and fight over the accessing of that memory.
With direct connect either CPU can access their own, or each other's memory without hits in performance.
' ">http://www.ZapWizard.comThe opinions expressed above do not represent the views of Advanced Micro Devices or any of their affiliates.