[citation][nom]BT[/nom]Why are they not using the 22nm process? Seems like that could save them on the power consumption and shrink the cpu? Is too expensive, and there fore cannot make money?[/citation]
22nm is probably limited in capacity because it is new and Intel probably doesn't have enough 22nm fabs to go around for multiple types of chips. We already have delays and such on Ivy Bridge so Intel decided to use fabs that they have a surplus of and were still good enough for their first Medfields anyway.
[citation][nom]Shmedfield[/nom]The fact that we only see such a select few Medfield benchmarks indicates that those are the only areas it does well in.If Medfield was so great, don't you think that the major international handset makers would all be lining up to be the first to release one? Instead, we're only allowed to see synthetic web browser performance, and we should assume from there that everything else about it is great, right?One must assumed that the very few handset makers that are releasing a Medfield phone were offered "incentives" to do so, 'cause the product obviously ain't selling itself.[/citation]
This is a CPU that is fairly incompatible with most of the software and such in the current market it is trying to get into so of course it isn't huge yet. Would you expect companies to line up for building machines with a 16 core ARM Cortex A15 desktop CPU if no software is around that can use it properly, despite it being pretty fast?
Honestly, just think about it and it should be obvious. The current Medfield cores seem to be almost as good as the projected performance of Cortex A15 cores so unless Intel starts ramping up the core counts and/or the 22nm node shows a huge improvement, ARM is still going to have faster CPUs for a while. At least Google is having a lot of the Android apps being eritten in code that isn't tied to ARM CPUs and shouldn't have a problem moving over to these Atoms without needing to be ported, but some Android apps use ARM specific code and obvioously won't work on Medfield and it's successors unless Intel does some hardware emulation (to save performance) and that seems unlikely because they would probably need to pay ARM for that.
Maybe Google can design a low overhead emulation for such apps, solving the problem for everyone themselves.