DSP Synergy.
The DSP announcement is nice because it opens the door to a whole new business that parallels WINDs current microprocessor business, one with equally radical growth rates. The DSP announcement is breathtaking because of the compounding effects of combinatorial mathematics.
To see the full implications of the announcement, begin by considering the following abstraction:
Every intelligent device of any interest to us is comprised of some combination of three primary types of processors: CPU, I/O Processor and DSP. A modern PC, for example, has one or two CPUs and perhaps any number of DSPs handling audio, graphics, etc. Tomorrow's PC will include one or more I2O chips for I/O processing. Any embedded device remotely interesting to us will be constructed around some combination of these three devices, many of which may only use one type, or any two types or, particularly in the future, all three.
Now, imagine you have an organization that designs and develops devices. If all you ever used were DSPs, then what are the odds that you will switch to WINDs DSP Tornado solution, assuming that WiSP has already been ported to most mainstream DSP processors? The only reason you wouldn't is that you are accustomed to the way you do development already, using a home-grown quasi-OS, or perhaps a rudimentary one supplied by the DSP manufacturer. So, the answer is probably significantly less than a one-half, say one in four.
If you only develop using a microprocessor, then the chances that you are using Tornado for VxWorks is at least one in three, assuming you are using a commercial RTOS; otherwise it might be as low as one in six.
If you only develop around intelligent I/O processing, then soon the chances will be effectively 100% that you will be using Tornado for IxWorks.
Now, over time your organization grows, expands or shifts its product line. Your product now uses both a MPU and a DSP, or a MPU and a I2O chip, or a DSP and a I2O chip. The only development environment that can be common between any of these combinations is Tornado. If your product involves all three types of processors, then not only is Tornado the only common environment, but it must be selected to keep the project manageable.
Thus, as combinations of types of processors need to be deployed in developing a product, only the MPU together with a DSP have a much of a realistic chance of escaping Tornado. A one in six chance combined with a one in four chance equals a three in eight chance. Only in this case is there much of a chance that Tornado will not be deployed, and even here WIND would quickly achieve a 37.5% market share.
Now start over. A year or two later the situation is much the same are described above, except jobs and people change. But along with these changes, WINDs market share for each processor in isolation is bounded by the worst-case scenario, namely 37.5%. This time the worst case chances of Tornado being deployed is 61.3%, and WIND totally dominates.
Of course there are flaws in the math, because there will always be pockets of development that stick for long periods to traditional ways using the same limited set of processor types. For example, VxWorks will not nudge Windows 95/NT from the desktop CPU - ever. But the point is that having Tornado importantly positioned on all three types of processors compounds over time, especially as devices become more complex and require more types of processors. This compounding will strengthen WIND's market penetration on each separate type of device, which in turn favors selection of Tornado due to further combinatorial effects.
Since WIND has a monopoly on I2O, if I2O becomes the de facto standard for virtually all I/O processing, then that monopoly alone would compound WIND's market share on MPUs greatly over time. Add in a significant DSP presence and the compounding will be breathtaking.
Allen |
