Completing the architecture, there are 4 units dedicated to the rasterizing operations, each one of them connected to a block of 4 SM, and 16 Polymorph Engines, the name chosen by nVIDIA To indicate the geometry units, each one of them integrated in a SM. For each GPU, nVIDIA has implemented 48 ROPs: comparing with the RV880 from ATI, used on the Radeon HD 5800 series, who has 32 ROPs, and the nVIDIA GeForce GTX 200 series, which had 24, it’s definitely an interesting increase.

It’s also interesting to notice how in different nVIDIA architectures, the number of stream processors are increased in proportion to the number of ROPs, confirming how important the shader elaboration capabilities are when it comes to rendering the most recent 3D scenes.