Fujitsu’s PRIMEPOWER UNIX® servers have obtained a leading market position for their excellent RAS(Reliability, Availability, Serviceability), not to mention their top-drawer processing performance. Now the SPARC64 V processors used in PRIMEPOWER have evolved once again through the adoption of brand-new technology. Here Katsuhisa Kubota, the enthusiastic general manager for next generation LSI development, introduces the distinctive features and attractions of the new product.

In order to improve system performance and reliability, processor developers come up with many good design ideas. However, unless they can be developed and implemented within the limited space available on the processor, they do not end up in the finished product. In other words, to achieve the desired results, specific design and manufacturing techniques are needed in addition to the design ideas.

In this second article, the spotlight is on “90nm process technology,” a manufacturing technology that achieves enterprise system performance and reliability. Introduced, among other things, is state of the art technology that overcomes problems in the development process plus enables achievements and business deployments for the future.

In the world of semiconductor production process refinement, there has been continuous research and development progress in making transistors and circuitry smaller. Currently the process technology used in most server processors is 130nm. But Fujitsu has announced a processor that employs next generation 90nm technology. This is the new SPARC64 V.

There are two main benefits from such refinement of process. One is the ability to place many more transistors in the same space (improved integration), and with such addition comes the possibility of adding a greater variety of functions.
For instance, the area of the CPU core and cache memory in the enhanced SPARC64 V has reduced with the change in manufacturing process from 130nm into 90nm. This made it possible to increase the amount of second cache memory from the previous 2MB to 3MB

Another is that such advanced miniaturization leads to direct performance improvement. Simply put, the faster the electrons move within the transistor from start to finish, the better the performance. Plus as the transistors reduce in size the resulting circuits are also smaller. This means, the overall distance to be moved shortens as the circuits shrink, allowing the electron to arrive at its destination earlier. I hope you can see this?

It is the road; however just simply employing 90nm technology doesn’t make it good. The specifications of 90nm technology differ depending on the semiconductor manufacturer. In general each new generation will improve efficiency on average by 25% (as was the case for instance, when process technology changed from 180nm to 130nm). But, when you compare the 90nm technology of other semiconductor manufacturers with the research and development of Fujitsu, our performance is another 25% faster. In other words, we believe we have achieved the efficiency improvement of two generations at a stroke.

As recently picked up by magazines and so on, leakage current has become a problem related to technology miniaturization. Leakage current is current that flows to places other than those expected. Therefore, even though circuits may not operate, electrical power is still consumed. But in comparison, we at Fujitsu have succeeded in suppressing leakage current to approximately 1/5th of that of general semiconductor manufacturers. This means that these advantages over other companies in performance and power consumption have been put to best use in SPAC64 V.

Table: Benchmark results for 90nm technology

On the following page the various technologies that enable 90nm are introduced.

Next