Mining ethereum – problem with R9 Nano cards – card is underperforming badly

First of all my setup:

Win 7 64
Asrock B85 anniversary
INTEL Core i3-4170 Processor
Intel ssd 320 series
8 gb – 2 x Crucial 4 GB DDR3 PC3-12800
PSU Thermaltake 1275W
USB risers 16x – 1x
Mining ethereum rig

And here is the problem:

My initial setup was 6 4Gb < PCI-E > DDR5 ASUS STRIX-R7370-DC2OC-4GD5-GAMING (RTL) DualDVI+HDMI+DP < RADEON R7 370 > and they worked perfectly, without any overclocking just factory presets they gave on average 12,3 Mh each (measured over 2 weeks). Each of those r7 370 has 1024 processors.
Now I decided to build another rig but more powerful. I chose R9 Nano, because of low power consumption, high amount of processors 4096, and moreover the price is cheaper as calculated per single processor. The card itself (R9 Nano) is 3 times more expensive then r7 370. So I thoughts, with 4 times the processors it should be hashing about 12*4=48 +/- Mh, so that should justify the cost + lower power consumption.

I bought 1 MSI card for test, plug it in and its total disaster, because without any overclocking it gives on average 20-24 Mh.
I tried overclocking it, it gives slight boost obviously, but not much. I tried changing every possible parameter I can reach and still can’t figure out what’s the problem. In theory, for this kind of mathematics (ethereum calculations I mean), it does not matter if its x1 or x16, so key factor at play should be GPU clock and number of cores. Its same clock as r7 370, but 4 times the cores, so why is it only giving me half the production?

Guys please help, if anyone had experience with this thing. What can I do actually fully utilize full card potential. Or otherwise maybe there is some reasoning why it will not give 4 times production vs r7 370? I can appreciate that relationship does not have to be linear here, i.e. more cores more Mh, but given that I at least expected to get 30 Mh +/-.
Advice, help, point in some direction where to research will be greatly appreciated!