IBM System z10 Microprocessor Chip
Through a great example of teaming, POWER6 engineers worked with z10 engineers to create the breakthrough z10 chip, which benefits from some shared technologies, but differentiates itself based upon a different design point. Some of the common DNA includes the IBM the 65nm silicon on insulator technology, and large portions of core pipeline design, the hardware decimal floating-point unit, and building blocks like latches, data flow elements, and SRAMs. (Note: for loads of detail, watch for future issues of the IBM Journal of Research and Development which already has a number of detailed papers in acceptance status for future publication.)
There are also some important differences, which gives the sibling chip its own very distinct personality. With more real estate given to caching (like about 20% more in a multi-tier configuration) and extra on-chip elements dedicated to sparing functions, the z10 extends the tradition of IBM Large Systems processors handling mixed workloads and larger working sets, while adding even more capabilities for CPU intensive workload. (Don’t forget the huge jump to 4.4 GHz quad core processor chips!)
In addition, there are the cryptographic and compression co- processors, Storage Control chips on each Multi-Chip Module, the new Symmetric Multi-processor topology, and larger caching- which includes support for the new 1 MB page frames for z/OS. Also, while the System z9 introduced hardware decimal floating point instructions in millicode, the z10 extends it to a hardware decimal floating point unit on each core. Along with changes in z/OS Language Environment designs, and backed by a new open standard definition for decimal floating point implementations, this increases accuracy, speed and capability for widely used commercial and financial applications and is an excellent example of systems design beyond just increasing the base chip speed.
So, like in any family, the z10 defines its own role in the family while leveraging the strengths and the lessons learned from its sibling. It is easy to just look at chip speed, but the detail of system design is what leads to balance systems throughput, resource optimization, and total value of the platform. There truly is nothing in the world like the z10 system platform, and it should be interesting to see the detail referenced above in those articles as they get published. I’ll keep an eye out for you!