Editorial: March 28, 1996
Time was, you built systems with tremendous capacities for future expansion and growth. Option slots, expansion boards, and physically separate subsystems linked by cables and connectors let you design for an uncertain future. However, designers often took one thing for granted: Fielded systems would be in the field for years.
No more.
Thanks in large part to rapid semiconductor improvements and the infectious Japanese approach to rapid product turnover in consumer electronics, life cycles of all electronic products have been sharply heading downward for years. If you design a system today, chances are good that in three years it will be irrecoverably obsolete. Consequently, you may now need to design just enough performance into your systems—enough to get your customers to the next all-new design.
PCs are everyone's favorite example of this trend. As little as two years ago, motherboard replacement was a good strategy for upgrading an aging PC. However, take a look at what's happened since then. The cost of hard drives has plummeted to below $300/Gbyte at the retail level. At the same time, average access times on these hard drives have improved by 30 to 50%. If you replace your motherboard, your old disk drive becomes the bottleneck.
The ISA bus, long the king of PC buses, has fallen, replaced first by the VL bus and then by PCI. Replace your motherboard with almost any sort of Pentium- or 586-based product and your video card, disk controller, and Ethernet adapter all become instantly obsolete. Newer systems require displays of increasingly high resolution. There goes your old CRT monitor.
There's no such thing as motherboard replacements for laptop computers.
Improvement through part replacement is becoming an increasingly nonviable strategy in PCs. The only real solution to upgrading any PC these days is to replace the entire unit. I believe that this situation also applies to many other markets. If you are using appreciable design effort and material budget on future expandability, I think you should reconsider. Unless your market still has long product-life cycles (many don't), you should take the "just-enough performance" route to stay competitive.
Steven H. Leibson
Editor In Chief