A Closer Look
NVIDIA's thermal solution is a complex system that consists of several components that can be removed one after another. First we removed the black plastic shroud, this is done by pushing in a small number of clips near the cooler base.
The main function of the plastic shroud is to direct airflow and to provide physical protection for the components on the card. The design also looks like something that could be easily customized by add-in board vendors.
Once the shroud is gone, you can take a close look at the actual heatsink design. A large piece of metal with five heatpipes soaks up the heat from the GPU core and transports it to the cooling fins where it is dissipated in the airflow. Please also note that this cooler uses a direct touch heatpipe design that we have first seen on CPU coolers.
Once the GPU heatsink is gone you are left with a black metal piece that is responsible for cooling the other components of the card. This is the configuration that many watercoolers or extreme coolers will use. If you take a look at the thermal pads in the third picture you can easily spot the cooled components: memory and voltage regulation circuitry.
NVIDIA is using a Delta fan, which is rated at 1.8 A, whereas most GeForce GTX 200 fans are rated at 1.4 A. This seems to suggest that NVIDIA designed their cooling solution with bigger heatloads in mind.
NVIDIA uses a 6+8 pin power design which is certainly needed. According to NVIDIA the max. board power of the card is 250 W - our readings say more like 320 W. Either way, a dual 6-pin design wouldn't be possible as that design is specified for only 225 W maximum power draw.
The GDDR5 memory chips are made by Samsung, and carry the model number K4G10325FE-HC04. They are specified to run at 1250 MHz (5000 MHz GDDR5 effective). This is the first time that NVIDIA is using GDDR5 memory on a high-end card, while ATI has been using GDDR5 for years now.
Hey, that's something new. NVIDIA uses a CHiL CHL 8266 voltage regulator on their card. To be honest I have never heard of a company by that name until I saw it used on the GeForce GTX 480. Unfortunately there is not a lot of info available about that voltage regulator, but it is certain that it supports voltage adjustments via I2C.
NVIDIA's GeForce 100 graphics processor is made on a 40 nm process at TSMC Taiwan. It uses approximately 3.2 billion transistors which makes it the most complex GPU built to-date. Please note that the silvery metal surface you see is the heatspreader of the GPU which measures 42.3 x 42.3 mm. The actual GPU die is sitting under the heatspreader, its dimensions are not known. NVIDIA did not communicate a die size measurement to the press.