Noise quality at high ISO is also determined by physics. The random "grainy" nature of 5D III ISO is primarily driven by the physical nature of light. It has far less to do with the electronic makeup of the sensor itself. Lower read noise might help the nature of noise at ISO 800, maybe ISO 1600 (MAYBE), but at high ISO, statistically speaking, the 5D II, 5D III, 1D X, D800, D600, etc. should all pretty much exhibit THE SAME kind of noise for an identical scene...as the hardware at ISO 1600+ really has very little to do with noise characteristics.
Thanks for the great explanation and insight, jrista! The only remark I have is that despite your theory that @iso1600+ these sensors should produce the same noise the raw samples I looked at show the more film-like pattern of the 5d3 vs. 5d2, an observation not only I have made?
But they have not yet put any commercial-grade high-volume sensor fabrication on those processes yet (god only knows why...if they don't do so soon, they will really be in a competitive bind.)
God might not know why, but marketing does - it's called "planned obsolescence" and means that you should time and stretch innovations so that users always find a reason to get a new product, either because the new one is better in some way (5d2->5d3: body, 5d3->5d4: sensor) or the old one simply breaks and repair is almost as expensive as a new product (most other electronics like lcd displays). Only fierce competition can prevent this method from getting too impudent - go Nikon, go!
Well, factually speaking planned obsolescence is really only an assumption and in some respects more of a myth (I don't know of any actual evidence that companies truly follow the concept of "planned obsolescence"...I've only ever known it to be more of a "conspiracy theory" kind of thing.) A company can also only really apply the notion of planned obsolescence in a market relatively free of competition. In a highly competitive environment, such as the digital camera market, conforming to the business ideal of planned obsolescence is more likely to make your involvement in that market segment obsolete, rather than simply your products. Canon is currently still the top DSLR dog, and in general the top digital camera dog as well, from a pure sales standpoint
so they still have some wiggle room left. From a technological standpoint
, almost EVERY SINGLE ONE of Canon's camera releases this year (except the 1D X) has been rather dull and lackluster, garnering a considerable amount low enthusiasm responses from their actual or potential customers, while competitors are wowing and wooing with some very impressive technology across the board (technical defects and issues, like the D800 AF and yellow LCD problems aside.)
Soon enough, though (few years probably at most), external competitive forces are going to do one of two things to Canon: Either force
them to innovate more and get ALL of the technology used in their product offerings "up to snuff", or cost them significant market share due to a generally inferior
product line. The kinds of improvements, such as low read noise, that have found their way into Sony's CMOS sensors are unlikely to stay there...they could be applied to, say, AF sensors, metering sensors, etc. (possibly allowing operation in EXTREMELY LOW light...say EV -5, -7?) It is also unlikely that we have seen the last of that scale of improvement in CMOS image sensor technology. BSI has been relegated to P&S cameras, but it has the potential of allowing near-micron pixel pitches even in FF sensors. High Refractive Index LightPipe technology can extend the life of FSI sensors down to around 2 micron pixel pitches. CP-ADC combined with active cooling could reduce read noise to fractional electron levels opening an even larger door for extreme DR (especially when combined with 16-bit ADC, we might see 15.9 stops of real-world photographic DR.)
There are a lot of improvements still to be made to digital image sensor technology. Some would say Canon is simply incapable of using a smaller fabrication process, and are stuck at 500nm without choice. I believe Canon has the ability to move off of a 500nm process to a modern 180nm process, although there are certainly logistical hurdles to overcome. I do not believe Canon can survive and try to follow a "planned obsolescence" model in the face of increasing and progressively intensifying competition from their competitors, though...so they will either have to stop prototyping
and start developing
technology, or eventually fade into oblivion. (Of course, the entire market segment might fade into oblivion if the worlds economic problems can't be reversed...and there doesn't seem to be any sign of hope on the horizon for the EU, China, nor the US.)