No actual figures yet, but I think the power consumption will be huge.
NVIDIA made its GeForce GTX 480M GPU official today. The DirectX 11 compliant GPU is based on the GF100 core and packs all the features of its desktop counterpart, such as decentralized hardware tessellation, next-generation CUDA and DirectCompute 5.0. The GF100 core has a configuration similar to the GeForce GTX 465 desktop GPU. It has three of its four graphics processing clusters (GPCs), and 11 out of 16 streaming multiprocessors (SMs) enabled, giving a CUDA core count of 352. To reduce the overall board footprint, the GPU makes do with a 256-bit wide GDDR5 memory interface, with 1 GB of memory. To make keep up with the electrical constraints of notebooks, the GTX 480M uses much lower clock-speeds than any desktop product that uses GF100. The core is clocked at 425 MHz, shader domain at 850 MHz, and memory at 600 MHz (real) or 2.40 GHz (effective), which gives a memory bandwidth of 76.8 GB/s. As mentioned earlier, the full feature-set of its desktop counterparts is packed with the GTX 480M, including support for NVIDIA 3D Vision, PureVideo HD, PhysX, and CUDA. It can pair with up to two boards of its kind in 2-way SLI. Constraints of the notebook form-factor won't allow any more boards, anyway. The GPU is open to Notebook manufacturers to plan their designs around. NVIDIA claims the GTX 480M to be the fastest notebook GPU. It finds direct competition in the ATI Mobility Radeon HD 5870, which is based on the 800 stream processor-laden Juniper core.