NVidia Launch GeForce GTX 660 OEM GPU
NVidia have silently launched the GeForce GTX 660 GPU for the OEM market. With 6 Streaming Multiprocessors (SMX), this card is aimed at the mainstream user, and operates at a slower speed than previous Kepler based mainstream and enthusiast class GPUs. As with other Kepler cards, it uses PCI-E 3.0 on platforms which support it, as well as having other features that were introduced with Kepler.
GeForce GTX 660 (OEM) has 1152 CUDA cores with a base clock speed of 823 MHz, and a boosted clock speed of 888 MHz. With 86 TMUs, the texture fill rate is 79 gigatexels/sec. The card has 1.5 GB or 3 GB GDDR5 memory, on a 192-bit memory interface. According to official specifications, GTX 660 OEM has a memory clock of 5.8 GHz (effective) and a memory bandwidth of 134 GB/s, although these figures suggest either a 5.58 GHz clock or a bandwidth of 139 GB/s.
Features include support for OpenGL 4.3, as well as NVidia standard features such as CUDA and 3D Vision. It also supports DirectX 11, PhysX, GPU Boost and Adaptive VSync, and both TXAA and FXAA. The reference design can use up to 4 displays, each with a maximum display resolution of 2560 x 1600 (digital) or 2048 x 1536 (VGA).
Standard display connectors include 1 x Dual Link DVI-I, 1 x Dual Link DVI-D, 1 x HDMI and 1 x DisplayPort. HDMI 1.4 and DP 1.2 are both included as standard, as is 7.1 channel audio over HDMI.
The standard card is 9.5 inches long, 4.376 inches high and two slots wide. The maximum GPU temp is 97C. It can draw up to 130W power, and uses a single 6-pin PCI-E power connector.