Strangely, after pocketing SGD$2498 Asus didn't bother to throw in any free games in the bundle (will check with Asus PR whether it qualifies for AMD's Never Settle promotion). Nevertheless the excellent GPU Tweak overclocking utility is included, and you can adjust the ARES II's vital clock and voltage parameters with ease, just like a regular AMD card (must remember to sync settings as each GPU is addressed independently). The GPU voltage can be adjusted up to 1.4V from the default 1.25v (no blackouts too during testing), but we cannot confirm its actual applied voltage from a physical multimeter readout.
Although the fans on the closed-loop liquid cooler cannot be controlled by software, it successfully tamed both GPU cores to an excellent result of between 54-56 degrees celcius on Furmark 1.10.4 full load, which is a far cry from the screaming inferno of the stock HD 7970 cooling solution. Coupled with a stronger set of fans, the ARES II might be able to perform even better from a thermal standpoint.
With slightly raised voltages, our review sample could attain 1250MHz on the core and 1800MHz on the memory base clock. As with any form of overclocking, your milege may vary.
Update 17/2: We had intended to capture and illustrate the frame time differences (smoothness/microstuttering) in a follow-up article delving into SLi/CrossfireX systems. Unfortunately FRAPS doesn't operate deep enough in the rendering pipeline to capture the combined frame times in a multi-GPU setup, leading to meaningless results. The alternative methodology would be a high speed motion camera setup/capture card to record the raw display output for comparison purposes, but unfortunately our lab is not set up for such testing at this time. In the meantime, here are some traditional fps-based benchmarks:
The ASUS ROG ARES II is (and will be for some time to come) the fastest AMD-based offering that money can buy, provided you are willing to sacrifice some degree of smoothness and driver-dependent game compatibility/rendering quirkiness that is inherent in multi-GPU setups.