Moore's Second Law and How the Itanium Architecture Suspends It

In 1994, Moore postulated a second law that qualified his first one. Interestingly enough, the law he cited was not so much technological in nature as economic. Although increasingly sophisticated tools can be used to cram ever more transistors onto a chip, the price of the manufacturing expertise rises accordingly.

Over the past three decades, the cost of a chip fabrication plant has soared so high as to make percentage comparisons meaningless. The plant is easily the most expensive facility that the industry has to purchase, especially taking environmental costs into consideration. In 1966, a plant might have cost $14 million, whereas a plant with similar output today has a sticker-shock price tag of one and a half billion dollars or more. This can be seen in Figure 2-2.

Therefore, the cost of increasing the number of transistors that can be put on the chip cannot help but increase even as the size of the circuits on the chip shrinks. Even though chip-making giants like Intel will be able to afford the more expensive and precise plants, the cost to the end users of the systems may increase faster than the increase in performance supplied by the new designs. The 'slack' in the equation so far is in the ability of the fabrication plants to make gains in both chip density, and in wafer output. At HP, we were seeing a slightly different effect of Moore's Second law.

With our RISC designs, typically, placing more transistors onto a chip did increase its abilities, but as we tried to add more capabilities to existing RISC designs, an incrementally greater percentage of the transistors became tied up in overhead to manage the added complexity of RISC. Design improvements such as out-of-order execution, which was very helpful in speeding up complex ERP applications, increased the overhead faster than the added performance. The billion-dollar fabrication plant may be placing a record number of transistors on the CPU chip. However, the net cost of the chips was increasing faster than the gain in power from the new designs.

This is where the Itanium processor arrives and suspends Moore's Second law. The Itanium processor family's unique architecture allows the processor to break away from the higher and higher percentages soaked up in overhead, allowing a performance gain that is unlike anything else in the market. Rather than spending to gain a top-notch fabrication plant for producing chips with moderate speed gains, the Itanium microprocessor takes advantage of both the speed gains from the manufacturing process and the next generation architecture to give superior performance for a much more attractive price.

It's this promise of lower price for superior processing power that will allow companies and research organizations to continue pursuing the trends of today's IT market. Higher levels of electronic commerce, handling streaming video, data mining, and complex problem solving that were once unheard of except on supercomputers will become possible in a much shorter period of time. The idea of computing as a service or as a utility becomes a reality, as computing can now be both more efficient and cost effective at the same time. We'll explore more of this emerging model in the next chapter.