The second-generation Fermi GPU is certainly more efficient, outperforming its nearest Nvidia counterpart, the GeForce GTX 465, almost everywhere on the board. What are the differences between the GF100 architecture of the GTX 465 and the GF104 architecture of the GTX 465? The GF104 has its components arranged differently and more simply (increasing yields), making it smaller, wider, and more efficient. It even sports a newer video processor, which is capable of something the GF100 GPUs aren’t: bitstreaming Dolby TrueHD and DTS-HD Master Audio soundtracks from Blu-ray movies.
While the GTX 460 has fewer shader cores than the GTX 465, it beats it regularly with its new resource arrangement. Now, instead of 32 CUDA cores per Streaming Multiprocessor (SM) that are present on the GF100 design, the GF104 sports 48 cores per SM. It has 4 dispatch units and 8 texture units per SM, compared to 2 dispatch units and 4 texture units on the GF100. The back end of the GF104 is also different from the GF100, and it has 4 ROP partitions instead of 6, which yield 32 pixels per clock and sport a 256-bit bus, instead of 48 pixels per clock and a 384-bit bus.
Cooler, quieter, and less power hungry, the GTX 460 is possibly Nvidia’s best “bang for your buck” GPU in years, not suffering the heat/power/noise problems of the first generation Fermi GPUs (GTX 480,470,465), and competing well against the HD 5830 and even coming close to the $290 HD 5850 at times. The $230 1GB GTX 460 outperforms the $200 HD 5830 on almost every benchmark that people put it up against, and, to sweeten the deal and lessen the $30 difference, users will get the advantage of Nvidia’s 3D Vision, PhysX, and CUDA technologies, at the minor opportunity cost of ATI’s Eyefinity technology. Check out the full specifications in comparison to the HD 5830, GTX 465, GTX 480, and HD 5850 below, and some benchmarks here.