This DirectCU II Top from ASUS has definitely been able to convince us. Thanks to the factory overclocking the card is eight percent faster than the reference clocked ones.
The cooling system proved to be very good too, it was almost noiseless in idle and managed to keep our card cool under heavy load remaining very silent which is very good. Furthermore we like the DirectCU II design.
ASUS completely revisited the PCB as well as the power design to offer a lot of electrical improvements over the reference board. Compared to the reference board (4+2 analog) the DirectCU II Top has a digital 8+2-phase Super Alloy Power design for exemple.
ASUS didn't overclock the memory although it wouldn't have been a big deal increasing the clock speed by atleast 50 MHz. It's not much but it would make perfect sense with such a premium product.

Thanks to the factory overclocking the DirectCU II is eight percent faster than the reference clocked GTX 680. Although there is a massive overlocking regarding the GPU it's not able to outperform the EVGA card we tested. The reason for that lies within the memory, since EVGA overclocks it and ASUS doesn't. Still the factory overclocking on the GPU is everything but shy. Another interesting fact is that our sample was working at 1'202 MHz all the time under load. It appeared to not make use of the boost feature. 1'202 MHz means that there is a factory overclocking of 144 MHz over the boost clock of reference cards.
When it comes to the cooler, this was the second best one we've ever tested on a GTX 680 that comes shipped with a card from the factory. After one hour of FurMark, the highest temperature measured was 62°C with the fans locked at 30 % (almost noiseless) which is excellent. At 50 % we measured 53°C max but in this case the noise level isn't good anymore. Indeed at 50 % fan speed the cooler was 5dB louder than the reference blower cooler. Be careful the cooler takes three slots, make sure you have enough space.
The power consumption isn't really good in idle as the DirectCU II Top was drawing 7 Watts more than the reference model. Under load it's better considering the performance ontop of the reference clocked card the power consumption was only 6 Watts higher.