A Closer Look
The DirectCU cooler uses five heatpipes that make direct contact with the GPU surface for maximum heat transfer. Two fans move heat through the fins to transport heat away.
Users of the first triple slot DirectCU designs were worried that their card might get damaged by the bending of the card under the cooler's weight. While there was never a real issue, ASUS has still listened to users and added strong support metal to the card, no more bending now.
The backplate is made of metal and provides additional protection against physical damage, it also might help a bit with cooling.
The card uses one 6-pin and one 8-pin PCI-Express power connector. This power configuration is good for up to 300 W of power draw.
ASUS has placed easy to use voltage measurement points near the right edge of the card (the three points on the left of the picture). Please note that the picture was taken with the PCI-E slot connector toward the bottom of the picture, so in a normal case the labels would appear to be upside down.
The ASUS GTX 680 DC II also features support for the VGA hotwire feature on ASUS motherboards. Basically it enables VGA voltage control from within the motherboard's BIOS for enthusiast modders. Using this feature requires you to connect the hotwires to the three points on the right of the first picture and enable the feature by bridging the two solder pads to the right of the transistor in the first picture. The resistors in the second picture are used to enable or disable certain protection features. For example PGR100 removes the 1.175 V GPU voltage limit.
For voltage control the card uses a rebranded CHiL 8318. It offers extensive software voltage control and monitoring features, but is quite new, so support in overclocking software is limited. ASUS' own VGA software supports it of course.
The GDDR5 memory chips are made by Hynix, and carry the model number H5GQ2H24MFR-R0C. They are specified to run at 1500 MHz (6000 MHz GDDR5 effective).
NVIDIA's new GK104 graphics processor introduces the company's brand-new Kepler architecture. It is NVIDIA's first chip to be produced on a 28 nanometer process, at TSMC Taiwan. The transistor count is 3.54 billion.