Is it just me, or are almost all of the microcontroolers in that "top 1000" list from Microchip? If so, that doesn't seem like a very representative sample.
There are other manufacturers in the top 1000 (e.g. Cypress, Adesto, Atmel) but currently the list does seem to be heavily weighted towards Microchip. This is a first stab at a component price index so it's a work in progress!
I don't really care about top microprocessors per se, I'd rather see a feature / price index. Meaning can I run a HLL on a micro, deep sleep, number of io, pwms, usb, etc.
Track mips/watt/$ and ram/$ along with forecast lines. If design with $3 part now, I'd like to know that it will drop to 1.75 in 2 years.
I find it rather ironic that filtering by price in product selectors is new and or rare. We exist in an economy predicated on cost efficiencies, where many of my designs are driven primarily on a price-point, part selection is an iterative process where cost absolutely plays a role and often the first one.
This sounds like very interesting analysis. You can actually get the data using the Octopart API (http://octopart.com/api). I'd love to see the results so please let me know if you do the analysis.
One then misses out on possibilities that open up by spending another 20 cents. Innovation occurs when new applications are unlocked by products moving into different economic regimes.
Boolean satisfiability allows me to meet a design criteria, derivatives allow me to innovate.
I would like to predict when full blown linux can run on a single chip (on die ram and flash) for less than $2 or below 10 mW. I'd like to be able to track when thresholds are crossed or a combination of features exist at a price point.
Really comes down to how Octopart determines relevance in the search. If "Relevance" is just another word for "arranged by sales volume", then it could be very representative.
It'd be okay if you created an index of the price of the top-selling N microcontrollers at any given time, rather than tracking them for their entire life cycle. The 8086 was a great and important processor, but it's no longer relevant to any discussion of processor prices.
The S&P 500 does this for stocks; perhaps their implementation and design choices can find some application here?
Yes, that would definitely be ok! The microcontrollers included in this index are a function how we rank them in the Octopart search index, which reflects some element of "top-selling" but probably not adequately so (i.e. your point on the 8086 is a good one).
Regardless, we think stock indexes are indeed are good role model for this kind of exercise. If there's interest in this we might invest in: (a) continuing to track this with better documentation on how we do it, and (b) start tracking more categories of parts.
"... but it's no longer relevant to any discussion of processor prices"
Kinda pedantic remark:
No longer relevant as a stand-alone device but don't overlook the IP core market. It's pretty common to find one or more of these old cores as a module in a more complex device (e.g. an FPGA) or as the base for an ASIC (esp. automotive). Same for the 8088, 8051, 6502, ...
Do such uses of IP manifest itself in this price index though? (I presume not, since even bare silicon prices should be several cents cheaper than a final packaged discrete device for SIP uses)
One extreme is ARM cores: incredibly popular, there are several in your laptop and probably half a dozen in your smartphone, ... ARM holdings is an IP & design company, they design & license cores.
But regarding those old cores I mentioned ... lots of use because it's legacy with lots of developers that know the architecture.
If following the S&P500, you'd want to weight it by the equivalent of "market cap". From what I can tell from the caption on the price index [1], this seems to be an equal weight index (which in the case of the stock market, results in a larger beta than the actual index).