Since the unveiling of the Fermi Architecture in the November issue of the NVIDIA Technology Conference, the point was that it can defeat HD 5000 series that ATI has dropped in the world just one week earlier, bringing along enormous results and the first DirectX 11 cards in the market? NVIDIA released TidBITS little here and there to show progress, all waiting for solid release date. Just after CES we had a chance to attend an in-depth look at the new architecture and see what it was really capable of. I got out of the eagerly awaiting the day when I could lay hands on the first 100 GF cards, now known as the GTX 480, to put it in the ring and see if they retained samples, we examined all she had, or if there was really no under hood.

Closer Look

It measures 10.5 cm long and has a full heat pipe stationed cooling solution to cool large die with three billion electronic transistors while working to keep fresh the memory 1536MB too. This card is PCI Express 2.0 NVIDIA and the expected response to the 5870 HD. Its best features include a total of three billion transistors used in 480 CUDA processing cores, a lot more than the dual 200 GT predecessors, but 32 below the archetype 512, we are shown, 60 texture units, 48 ROPs, and 15 streaming multiprocessors 4 groups GPU. There are a total of 1536MB of GDDR5 memory running through a 384-bit bus, which combine to provide a memory bandwidth higher.


Connectivity is through a pair of dual-link DVI and a mini HDMI port that is able to produce high definition audio. The connections of power to the GTX 480 include two types of input methods. A single 6-pin connection and a single power 8-pin PCI-E from which either can be used. This card is able to use an SLI bridge connection SLI. SLI expansion, it is said in the percentage 80 + with two cards, but adds a third and calibration must be pretty impressive. The cooling system of the GTX 480 is composed of a base heat pipe system connected to an array of fins under the metal cover that helps bring more heat. The four heat pipes sit directly above the base GTX 480 which is not the motherboard to create a barrier to dissipating heat load must be somewhere in the range of 250 watts maximum TDP. The air passes through the housing and the exhaust heat out of the box.


480 GF GTX 100’s architecture is built on Fermi and performs at frequencies of core and 1401MHz shader, 700MHz core and 924MHz GDDR5 memory is. Frame buffer memory is 1536MB 384-bit bus passes and port 177.4 GB / s memory bandwidth. GPU consists of four groups that host the 480th core CUDA, 60 texture units, 15 streaming multiprocessors and 48 of the SPD. For memory, NVIDIA has decided to use the modules from Samsung.