The marketing of modern handheld chipsets

Many great things start with an idea. In the case of most handheld chipsets out there, though, that's probably not the case. It's much more likely your feature list will be dictated primarily by your marketing department, and then the engineers will try to figure out the best way to manage that in the target timeframe. They'll talk with marketing until they (hopefully) come up with an overall design that makes sense in terms of both costs and market appeal.

That doesn't mean there aren't killer ideas some engineer came up with out there, but this kind of thing tends to be restricted to startups or small companies, for obvious reasons. A good modern example of such a thing is Bitboys' Vector Graphics core, the G10, which was so small they could integrate it directly onto one of NEC's LCDs, rather than as a separate chip. This even saved a fair bit of power and, needless to say, Nokia loved them for it.

Of course, this kind of dynamic is true of much of the consumer electronics market, up to a certain extent. In the PC CPU space though, to take a relatively simple example, marketing's influence is relatively much more limited. It's not your marketing department that's going to figure out how to maximize performance-per-dollar.

But on the other hand, they'd still be the ones to decide how much of a focus on performance-per-watt there should be, for example. One of the rare cases where marketing apparently got significantly more involved is Intel's Netburst architecture, used various processor ranges from Celeron to Xeon and Pentium in between. It was focused nearly exclusively on clock rates and not actual performance, which resulted in a competitive disadvantage against AMD.

Furthermore, just like in many other consumer electronics devices, handheld chipset manufacturers have very little say on the final product that will get sold to the end-users. Most of the time, off-the-shelf designs are marketed to a wide range of potential customers, and generally the most cost-competitive designs for a given level of market differentiation win the sockets.

There's another (but much rarer) alternative, though, which is when large handheld manufacturers such as Apple or Nokia go straight to potential partners and tell them more precisely what they need, and in what timeframes. In these cases, it's obviously much easier to come up with a list of requirements for your engineering team; just copy-paste what your potential customer told you they needed! Custom chips are also a possibility, but these are much less frequent even then because they'd affect time to market, and have large one-time design and tape-out costs.

If you go for a custom chip, only two factors truly matter anymore: your architecture's efficiency and company politics. Albeit not really custom designs, there's an excellent (albeit only rumoured) example of the influence of politics with the iPod Nano socket PortalPlayer lost to Samsung earlier this year. Apple is one of the biggest consumers of NAND Flash memory in the world, and Samsung is one of the biggest manufacturers of NAND Flash.

But Samsung also produces a variety of other chips, including application processors. And they use their own fabs, which gives them an obvious cost advantage. Furthermore, Apple always fears being NAND Flash supply-limited in the longer term, so they tend to do whatever is necessary to secure capacity. If you combine all of these factors, it didn't really make sense for Apple to give the deal to anyone else, especially so when PortalPlayer's technology wasn't even more feature-rich than Samsung's.

This leads us to the next part of our little analysis. If we consider the preceding factors to be negligible in a specific instance, we should then focus on the actual quality of the various hardware solutions. And assuming the same manufacturing partners, which is often true here, that means architectural efficiency - so, what's efficient?

Stay tuned for part two where we explain how efficiency is achieved.