The Leapfrog Effect

"We follow the industrialized countries, especially USA, so we can choose the winners and avoid the expensive market experiments …There is a leapfrog effect, given that a lot of companies are just beginning to manage their workflow through Intranets, or even their connections with providers and/or distribution channels with extranets, and they don't have legacy systems that slows introduction of new technology."

-Anthony Rodriguez Chapa (Chapa, 1998)

In the developed world, we are inured to rapid change in IT. Advances tend to be made in an incremental fashion, with one small refinement made on a chain of backwards-compatible products. The cost of following this path, in terms of change management, administration and training, is significant.

Developing countries simply cannot sustain progress along this incremental path. However, due to the lack of investment in legacy systems, hardware and software, they can be in a good position to "leapfrog" over some of the incremental steps and to select a new position on the technology curve, as noted above.

The precise nature of this "leapfrog effect," particularly in relation to sustainable development, is worthy of further study. In this regard, it is worth noting that donations of outdated or discarded equipment from developed nations are often unsuccessful as they limit or prevent the occurrence of this effect.

Leapfrogging Moore's Law

In 1965, then-Chairman of Intel, Gordon Moore, made the prediction that the number of components on a microprocessor integrated circuit roughly doubled every year. In 1975, he revised this prediction to the number of transistors doubling roughly every two years. This prediction has held true against the empirical evidence from 1971, right through to the current generation of Pentium processors offered by Intel.

Figure 1 shows the increase in the number of components in microprocessor integrated circuits for a period of some 30 years. The actual trend is very close to that what was predicted, with the number of transistors on microprocessor devices doubling roughly every two years. However, there are some aspects of this graph that deserve comment.

Figure 1: Actual and projected trends of computing power from Moore's Law ^[1]

If we focus on the period between the introduction of the original 60-MHz Pentium processor in 1993 and the introduction of the 1.5GHz Pentium 4 in 2000, we can observe that:

1. The actual increase in numbers of transistors fell below the projected trend.

2. The number of different microprocessors introduced over this seven -year period (13) was greater than the number of processors introduced in the previous 11 years (8).

If we pick the points on the curve in this period where the projected trend was closely approached or exceeded, we can identify "leapfrog points". These are points at which there is a suitable technology whose capabilities are not significantly exceeded by the next incremental offering.

1. The original 60-MHz Pentium introduced in 1993 (60 MIPS).

2. The 300-MHz Pentium II introduced in 1997 (300 MIPS).

3. The 1.5GHz Pentium 4 introduced in 2000 (1500 MIPS).

These leapfrog points represent a five-fold increase in computing power over a four - year cycle. For an organization depreciating capital expenditure on Information Technology over the same period, making purchasing decisions at these points would be appropriate.

This period where the trend drops below that predicted corresponds to the widespread uptake of personal computers, especially those adopting Microsoft Windows as the operating system. The larger variety of processors introduced during this period corresponds to the demand from different markets (e.g., servers, desktops, laptops, embedded computing) that exploded at this time.

Some of the processors offered were clearly, what might be termed "market experiments" and some failed. In particular, the Pentium III was consistently underpowered in all varieties and provided no real incentive for an end-user to upgrade hardware.

For developing countries, not only is there no incentive to upgrade, there is no capital available with which to do so. A common scenario is for an organization - commercial or educational - to struggle along with whatever equipment they have until it fails totally and then to install the current best hardware. This corresponds to leapfrog along the technology curve and the effects can be quite dramatic ^[2].

Moore's Law is of interest because it is almost unique in industry. No other technology save telecommunications exhibits such a dramatic increase in capability over a short period. The consistency of Moore's Law is due to a number of factors:

• Technological control of a physical process (etching of semiconductor substrates by various photo-lithographic processes).

• Incremental refinement of the manufacturing process through experience.

• Improvement on a single aspect of the manufacturing process yielding improved performance (e.g., by decreasing the feature size, more transistors can be packed on the same area, providing a more complex processor).

In fact, experience gained through the manufacturing process and the volume of manufacture lead to cost reductions over time. As the power of processors has increased dramatically, the average cost of a personal computer has reduced over the same period. As personal computers crossed the threshold of affordability in the developed world, demand increased further, permitting greater volume of manufacture. In fact, a positive feedback loop drives Moore's Law. The feedback forces the semiconductor industry to keep pace by introducing new fabrication techniques and continually reducing costs. At least until the physical limits are reached, the consistency of the law for the foreseeable future is assured.

Since an organization can rely on the consistency of the law (projected out until 2007, it still holds), they can make decisions about depreciation, purchasing and/or proposals for technological aid. They are able to hold position on the technology curve without fear of being left dramatically behind; knowing that a suitable landing spot for the leapfrog jump will appear in due course.

The implication of Moore's Law is useful for technology development projects that may be spread over several years. It makes sense to specify allocations of funds for computer hardware, rather than specifying particular technologies at the proposal stage. Purchasing decisions should be delayed as long as possible, leaving only sufficient time for procurement, delivery and installation. Clear policy decisions should be set out in funding proposals for replacement of hardware after specific periods of time.

Microprocessor development over the past 30 years has striven to maintain backward compatibility. Indeed, only the 2001 release of Intel's Itanium processor breaks the link of compatibility that extends from the Pentium back to the 4004. In part, this is due to the incremental nature of semiconductor manufacture - using past experience to drive cost-reductions and new technologies. In part it is due to the need to protect the investment that consumers have made in the technology. This is in sharp contrast to many of the software applications that have been developed throughout the personal computer revolution.

Leapfrogging Software - The Arms Race

The advantage that hardware has over software (at least for the manufacturer) is that it wears out, develops faults and eventually becomes obsolete. No matter how reliable we think our computing equipment might be, it is still prone to catastrophic failure, loss or physical damage. Software has no such characteristic - if software is functioning at a particular time, performing some function, it will continue to perform that function at some other time. The set of instructions that software follows can be regenerated by compilation, or copied perfectly as a collection of bits. Software manufacturers therefore have a problem - if they release a single version of a software product, then the market will eventually reach saturation. After an initial peak, sales will gradually drop as new computer users adopt the product. There is limited potential for cash-flow. Software manufacturers therefore must continually provide new releases or new versions of their software in order to protect their source of income.

In a very real sense, there is an "arms race" (Thimbleby, 2001) between users and software vendors. The vendor has limited interest in providing backward-compatibility between software versions since they wish to drive demand for the latest version. As users adopt the later versions, they find it more and more difficult to exchange information with those who still maintain the earlier versions.

This kind of software incompatibility is a major barrier, especially for developing nations who are more likely to lag behind on the technology curve. They are forced to upgrade software in order to communicate with others, thus they are forced onto the incremental path. Worse still, if a software upgrade requires a hardware upgrade for which there is no budget, an organization can find itself left further and further behind.

This software barrier to the leapfrog effect goes some way to explaining why donations of older computing equipment from developed to developing nations are often unsuccessful. The older hardware immediately sets the recipient years backwards on the technology curve and since it is unlikely to support the newer software, cuts them off even from the incremental upgrade path.

^[1]Source: Intel Inc.

^[2]This author has observed a leapfrog of 13 years (upgrade from 80386 to Pentium II hardware) in an educational establishment in Papua New Guinea. The author is currently using hardware and software that is five years and three processor generations behind the cutting-edge.