According to previous updates, the production chips would have have greater efficiency:
Normal Mode:
0.7v vdd
6gh/s per chip
0.343w/g on chip
And looks like real performance of samples is better than predicted:
...
6.4GH/s | 0.2363W/G
6.8GH/s | 0.2439W/G
7.2GH/s | 0.2495W/G
...
while the power numbers are certainly good, there are two questions i'd have.
the main goal when doing bitcoin mining is balancing the capital cost and the operational cost.
the capital cost can to a large extent by expressed as a performance/density figure, GH/MM, and the operational cost expressed as a Power figure W/GH/S (aka J/GH). the remaining system costs (fans, power supplies, pcbs etc.. are largely constant regardless of the silicon used, so the main factor is efficiency and cost of the silicon itself)
clearly, many designers are working on asics that run at low voltages and thus achieve a headline low power figure.
The key is to achieve LOW POWER, while at the same time, ensuring that the silicon cost is low.
So it looks like the nominal performance of the be300 chip is somewhere between 3 and 6 GH/s (or 2.8 to 7.2 gh/s) per package.
And we've been told the package size is a 5x5mm square. Since we havnt been told the die size... lets try and guess. clearly its nowhere near 5x5 mm... so perhaps its 3x3 (=9 mm) or 4x4 (=16 mm)?
it makes a difference how big the die area is, as the gigahashes per mm need to be significantly better than last year's asics if its to be competitive in 2015. it needs to be above 3 GH/mm to be competitively priced, and ideally north of 5 gh/mm. And if... completely guessing... the die is... lets say its a 2.5 x 2.5mm, (= 6.25mm) then its going to be lower than 1 GH/mm, and thus will be expensive compared to other competitors coming out in 2015. Bear in mind that last year's chips were all above 1 gh/mm, also in the tsmc 28nm process. this year's ones should be significantly more efficient in order to bring down the cost of mining.