Originally posted by reh321 If Pentax
does depend on the PRIME processor to perform AF, that could provide a major explanation for why their cameras lag behind Canikon in AF - from the reading on-line I've been doing, for example
dedicated auto focus processor: Canon EOS-1D / 5D / 6D Talk Forum: Digital Photography Review
there seems to a general understanding that Canikon uses a dedicated processor in most cases. A DSP {"digital signal processor"} is especially designed and optimized to do certain tasks like this, and in the hands of a skilled developer will always come out ahead of a general purpose processor of comparable technology.
It all isnt a simple connection.
The 6D has only a single DSP, same as the 5D3. The 6D autofocus is inferior to a K-3II autofocus in all aspects of precision, speed, tracking, darkness (
using comparable new glass; we just had a little fun test under colleagues), while this is the other way around for a 5D3.
The M5 has a single DSP two generations newer than a 5D3 and still is not match.
What is quite visible in the viewfinder and with reactions of the lens when using "auto" type AF modes where the camera is supposed to track a subject across all available AF points when the photographer is not tracking properly is the AF indicators turning on and off all the time. But the frequency of this is much higher on the Canon still.
So either the PRIME hardware is slower than the DIGIC or they do not dedicate similar computing power to AF routines versus other things in software or the algorithms used are slower. Obviously the faster you can compute and adjust to changes the better the results will be.
The other unknown factor is what the AF hardware (AF lenses and AF pixel sensors) makes for a difference. Double cross type sensors do provide twice the data to find the right focus. Nobody knows if Canon hardware sensors has more pixels or more sensitive pixels or whatever.
On the other side I am rather sure that the sheer number of AF points created in software from the AF hardware pixel rows has a very minor effect on most results.
Lastly the image recognition speed and intelligence used on the 360x240 RGB pixels could have also major impact. If you do that correctly 80% of all tracking is based on this and not of the very, very limited information from a few color blind AF pixel rows.
But: Nobody outside the manufacturers has any clue.