Zertz wrote:You are entirely right, if we look strictly at load numbers. The thing is, those UMPC's almost always idle or do something easy for the processor (word, internet, etc) and both CPU's take roughly the same power in those conditions.
I haven't seen the actual data to confirm whether the Nano has similar power at idle and light loads, but even if you're right, you can't design a UMPC around the average power of a processor. You still have to account for the power under load, which affects the design of the cooling and thermal equipment, the form factor, and the air flow. Otherwise, your UMPC melts into a puddle of molten plastic and trace metals the instant you run a high power app.
Zertz wrote:Even at full load, Nano takes almost 50% more power, but it gets the work done 30% faster!
This is the case when Nano is operating at 1.8GHz, but at that speed, the peak power dissipation is close to 25W. In order to run the processor at a thermal point suitable for a UMPC, you'd have to use the 1.0GHz Nano @ 5.0W TDP. At this point, you're losing 40% or more of your performance ( 1.0/1.8 ), and the performance and power efficiency of Atom starts to look pretty attractive.
This was the point I was trying to get at. On the desktop, assuming you can find a Nano at a price that's comparable to Atom, Nano is probably the better deal. In mobile, and especially in the ultra-mobile space, Nano can't get down to the power levels of Atom, and if it tries to get close, it ends up losing all its performance.
Atom still runs at 1.6GHz @ 2.5W of peak power, in the case of the Atom N270 mobile sku. But Nano has to downclock all the way down to 1.0GHz just to run at 5.0W of peak power. Atom also uses a lower power chipset in mobile, which decreases the power versus Nano even further.
Of course, it will be interesting to see how the power affects actual usage of devices built with Atom and Nano, assuming LR or other reviewers find devices with either processor and do some actual benchmarking with battery testing.