NVidia Launch GeForce GTX 660 and GTX 650 Graphics Cards

NVidia have launched two new graphics cards, the GeForce GTX 660 and the GTX 650. The GTX 650 uses GK107, the die introduced with the GeForce GT 640 cards, and has 2 Streaming Multiprocessors (SMX) with a total of 384 CUDA cores. GTX 660 cards use a new die, GK106, which has 5 SMX for a total of 960 CUDA cores.

GeForce GTX 660 runs with a base clock speed of 980 MHz, which increases to 1033 MHz when boosted. With 80 texture units, it has a texture fill rate of 78.4 gigatexels/s. 20 ROPs provide a pixels throughput of 19.6 gigapixels/s. The lower level card, GTX 650, has 32 texture units, producing 33.86 gigatexels/s. It's 8 ROPs give 8.46 gigapixels per second.

GTX 660 has 2 GB GDDR5 memory on a 192-bit bus, running at 1.5 GHz (6 GHz effective) and with a bandwidth of 144 GB/s. The slower GTX 650 has 1 GB GDDR5, clocked at 1.25 GHz (5 GHz effective) on a 128-bit bus, and it has a bandwidth of 80 GB/s.

Both cards have all the usual Kepler features, including OpenGL 4.3 and DirectX 11 support. In supporting systems, it also uses PCI-E 3.0, providing extra benefits in some cases. Both cards have a dual-link DVI-I and dual-link DVI-D ports. GTX additionally has an HDMI and a DisplayPort outputs. GTX 550 has a mini-HDMI port.

GTX 660 is 9.5 inches (241 mm) long, and GTX 650 is 5.7 inches (145 mm) long. Both cards use two slots, and have a standard height of 4.38 inches (111 mm). GTX 660 is rated to use up to 150W, compared to GTX 650 at 64W - although the latter still needs a 6-pin power connector.

GTX 660 is available now at many retailers, for a price of $229. GTX 650 is priced at $109.