|
|
|
|
|
|
|
interface quickly becomes the system bottleneck. Higher-bandwidth implementations such as crossbar switches would be required to force down contention. |
|
|
|
|
|
|
|
|
There is no conclusion to the design processmerely a termination. Design iteration follows iteration, each improving our understanding of the design tradeoffs of cost and performance. At some point, the design is "frozen" so that implementation can be completed. At that point, the designer can only hope that the component parameters, which are changing as time goes by, are correctly predicted. Mispredicted parameters, overlooked functional assumptions or interactions, flaws, etc., create sleepless nights for the designer. Each is repaired or "featured" (a feature is sometimes a flaw that we learn to live with), hopefully with minimum impact on the implementation. |
|
|
|
|
|
|
|
|
At the end, the design team, having learned the lessons for the current project, moves on to the next design which has new technology, marketing, cost, etc., assumptions. The rapid change in the computer design field diminishes, but does not negate, the value of "the last project's" lessons. In each case, however, the better the understanding and the better the analysis, the better the resulting design. |
|
|
|
|
|