Next Up Previous Contents Index
Next: Summary of recommendations Up: Autodesk confronts an Previous: AutoCAD as an

Anticipate technological change

There's one good thing you can say about recessions. They sure do make hardware cheap! The 1974-75 recession was a major contributor to the subsequent personal computer explosion in a very simple yet often overlooked way: at the end of that downturn, the price of logic circuits per gate had dropped from dollars to pennies. Projects that previously required the financial resources of a substantial company were now within the purview of individual mad scientists in drafty basements. I know. I was one.

The semiconductor business is an analytical economist's dream. It has a huge capital spending component and long lead times, a large intellectual capital contribution that's related to R&D spending and the ability to translate it into marketable products, and close to zero material costs. These fundamentals have interesting consequences in a recession.

The R&D costs for the products you're currently making are already sunk and can be recovered only by selling more products. The cost of operating the production line is dominated by amortisation of capital equipment already in place and direct labour--both unrelated to the volume of products produced. Incremental improvements in yield reduce costs and are sought on that basis alone. The cumulative effect of these fundamentals is very much like a replicator that's swallowed its shut-off switch. With the factory accounting for a constant fixed cost every month, all incentives are toward increasing output. Increased output drives the unit price down, forcing competitors to further streamline their own operations, and so on.

The nature of semiconductor fabrication technology causes memory circuits to lead all others in complexity at constant cost. A software company whose products have been constrained by limited memory for the better part of a decade would be wise to ponder the consequences of a memory price crash. (Or, should no recession eventuate, a slower yet equally inevitable erosion of memory prices.)

When Autodesk was founded, 64K memory was the standard complement in the personal computers of the time. I recall when I first heard that the IBM PC version of AutoCAD would require 128K of RAM. ``How profligate,'' I thought, ``Why can't they fit it into 64K like real programmers?''

At that time, 16K RAM chips were giving way to the 64K RAM generation. Since then, we've seen the 64K RAM supplanted by the 256K RAM, and the 256K pass the torch to the 1 Mb chip. Recently, 4 Mb RAM chips have begun to appear in quantity with prototypes of its successors, the 16 and 64 Mb chips already described in research papers at the various solid state circuit conferences.

As we sit and struggle with MS-DOS and its one megabyte address space limitation, it's worth contemplating the implications of a world that's adopted the 4 Mb or 16 Mb RAM as the standard memory component. Most of the microprocessors in use today have a 32 bit memory bus: each memory access reads or writes four bytes at once. Hence, when that memory is composed of chips of a given capacity measured in bits, the memory expansion increment in bytes is four times the chip capacity in bits.

Think about it. As 4 Mb RAM chips achieve price parity and begin to displace the 1 Mb generation, the minimum memory configuration of a 32 bit bus processor such as an 80386 or 80486 will be 16 megabytes, and memory expansion will be in increments of 16 megabytes. When the 16 megabyte chips supplant the 4 Mb chips, memory will start at 64 megabytes and grow in steps of that size.

Memory on this scale truly changes everything. CAD moves from an application that strains the limit of every resource on the system to a modest user of abundant computing resources. Anticipating the widespread availability of computing power in this class (which will happen regardless of the economic environment, but perhaps later rather than sooner), Autodesk should devote resources within the Advanced Technology division to defining the design tools made possible by such hardware configurations and undertaking their development in order to deliver them as soon as the required hardware becomes available to our customers.


Next Up Previous Contents Index
Next: Summary of recommendations Up: Autodesk confronts an Previous: AutoCAD as an

Editor: John Walker