While there are still a few hours left to the official announcement, many details of the "fastest DX11 GPU on the planet" seem to have been revealed by various partners, including MainGear, CyberPower, GigaByte, and EVGA. Nvidia’s latest GPU will indeed be called the GeForce GTX 580, and is said to offer up to a 35 percent improvement in performance over the inimitable GTX 480. The GTX 580 has a 772MHz core clock speed, 512 CUDA processing cores, and a 192.4GBps memory bandwidth powering 1.5GB of 384-bit GDDR5 RAM, which is effectively clocked at 4GHz.
Other details have been by CRN, indicating a price of $499, three-way SLI support, 10.5-inch length, 6- or 8-pin power connectors, and 16 polymorph engines that are capable of processing 2 billion triangles per second, for the best tessellation performance. The GTX 580 also features a built-in hardware monitoring system that will also allow the user to dynamically adjust performance along each 12V rail.
[RELATED_ARTICLE]Benchmarks are forthcoming, and it’s when we’ll see how Nvidia's GF110 Fermi architecture - and all those CUDA cores, large memory bandwidth and GDDR5 RAM - does against the HD 5970, GTX 480, and HD 5870. Things are sure getting hot, and AMD’s riposte is expected soon enough, with the high-end performance HD 6900 series, which includes the proposed the dual-GPU Antilles HD 6990. Calling the GTX 580 “Swift. Stealthy. Deadly.”, Nvidia sure wants you to know that its flagship GPU is quieter and less power hungry than its predecessor, and you can see some of its power/consumption stats in our previous coverage, in the technical demonstration video highlighting its tessellation and COD: Black Ops performance