SAMA5D3 is 65nm and does 536MHz at less than 200mW for the whole SoC. Sitara is a 1GHz Cortex-A8 at 65nm as well and Vybrid uses a 40nm process. However a 486 is not in the same league as these Cortex-A5/A8 based SoCs, so Quark would need to sell significantly below the $7 Atmel asks for SAMA5D3.
Yes, there are claims of a "Pentium" instruction set but it is literally a 486 with a 5-stage single-issue pipeline and 16KB shared I&D cache (compare the timings with an old 486 manual or look at page 20 in the last link). Remember the current Atom is already Pentium class, so to get x86 any smaller you have got to get down to a 486! The die is about 30mm^2 based on the photo in this article. TDP is 1.9-2.2W for the SoC.
Forget FFTs, Quark won't ever run any DSP code. Multiply takes 6 cycles vs 1 cycle for a MAC on Cortex-M3/M4. Floating point is even slower: fadd is 10 cycles, fmul is 11 - compare that with 1 cycle for fadd and fmul on Cortex-M4...
It is good Intel couldnot or didnit enter smartphone business. They are experts in making processors and it would be good if tghey keep that monopoly going. Just joining the smartphone wagon would not earn any good.
Yup, I forgot about the A13, which is sold in small quantities by Olimex. I wouldn't be surpised if the packaging impacts the performance. Also, it's very tablet oriented, not industrial, and I wonder how long it will be in production.
Another point about the Quark: availability and price from Mouser & Digikey will be very important if Intel wants to be serious about embedded again.