Nano crushes Atom

[JumpingJack]

Very interesting power consumption comparison indeed.

Yeah.... and completely wrong. They are using high power rated PSUs for ultra low power boards. At < 20% loading, a 500W PSU drop efficiency like a brik... some to below 50% efficiency.

If they did the measurement right, the Atom board can be shown to consume no more than 34-35 Watts with drives and with a power delta of no more than 2 Watts idle to load.

All the sites so far that have published numbers have failed to specify what PSU they used. At these loads, it is impossible to draw correct conclusions when your 20-30W board is being driven by a massive power supply.

To demonstrate this I profiled the same board in these reviews using a CoolerMaster 500W RP-500-PCAR power supply, I used OCCT 2.0 with a CPU only selection and a 20 min custom duty cycle. I then compared the exact same board, just switched out PSUs and used a Sparkle 220 W 80+ efficiency PSU.

As you can see there is about a 20 W difference between the two traces, i.e. PCPer, Hothardware, and Ars have at least a 40% error in their measurements. In the case of the coolermaster PSU idle was 55.3 and on the Sparkle Idle was 34.0. Basically, I can reproduce the 55is-60is Watt numbers with easy, but only when using the worst possible choice of PSU for the job.

From the sparkle data I can derive a better approximation of the CPU loading power from idle to peak, http://www.silentpcreview.com/article773-page4.html . I am measuing 2.1 W idle to peak, and accounting for the efficiency at the input power around 34 W, the actual CPU consumption is around 1.6-1.7 W well within the TDP (as expected).

2 watts out 60 vs 2 watts out 34 is a huge difference, and the Atom tests are being dominated by the cruddy choice of PSUs. Taking this out, Atom take a strong lead in perf/Watt (I suspect). I would need to get my hands on a Nano board/CPU to do the same tests.

EDIT: BTW power is recorded on a data logger, a WattsUP Pro ES.

The review sites are getting sloppy.... very sloppy.

Jack