|
|
|
|
|
|
|
Consider the issue of compatibility. Companies such as IBM, DEC, Intel, and Motorola have generally retained design features, instructions, and even data types that may have effectively been made obsolete by the introduction of more powerful or more robust design attributes. Older instruction types and data types are retained to preserve the portability of programs to the newer processor. This enhances the user acceptability of the design and lowers user cost in adopting the new product. Similarly, the notion of an architecture family is very important in design, sales, and marketing so that a product may be extended later as needs arise to include the possibility of enhanced performance. Yet upgradability as well as compatibility have distinct design costs, as provisions must be made for each design feature. Each feature included affects design time and time to market, whether directly as part of the principal design attributes or indirectly to support compatibility, upgradability, serviceability, or other requirements. Design time represents a fixed cost and a call on design resources, as well as possibly delaying access to a particular market. The more complex the design and the more attributes to be considered, the longer the design cycle and the more design resources and fixed cost required. |
|
|
|
|
|
|
|
|
Let us explore the costs in more detail. There are two types of cost: fixed cost and the marginal cost of manufacture. Most designers, in the absence of other information, think of cost in terms of the marginal cost to manufacture the next processing unit. The fixed cost includes the engineering cost of both hardware and software, engineering charges for building VLSI masks, charges for both CAD equipment and software, OS and computer software development, and manufacturing tooling expense. On top of this is an allocation for general management and overhead, as well as for marketing development effort. Once the product has been developed, manufacturing commences; but the cost of manufacturing per unit is determined by a learning curve. Early units are necessarily more expensive than later units, both to build and to maintain. After a while the manufacturing process itself becomes more tuned to a particular product and becomes more efficient at producing and testing the product. |
|
|
|
|
|
|
|
|
Finally, the marginal cost to manufacture is still only one component in the marginal cost to deliver a product. The cost to deliver includes transportation, sales, general marketing, warranty service, and documentation costs. |
|
|
|
|
|
|
|
|
A processor designer generally has little control over many of the aforementioned fixed or variable costs. It is largely in the area of marginal manufacturing costs that design tradeoff is possible. From time to time throughout this text, when we discuss the issue of cost-performance, it is in this narrow marginal sense. The designer must be aware that the marginal cost-performance we address here cannot be done at the expense of a significant side effect on either the fixed costs or the other variable costs. For example, a significant increase in the complexity of a design may directly affect its serviceability or its documentation costs, or even the hardware development effort and time to market. These effects must be kept in mind, even when it is not possible to accurately quantify the extent of the particular side effects. |
|
|
|
|
|