Introduction

Close to ten months after kicking off its DirectX 11 compliant Fermi GPU architecture, NVIDIA carefully treaded lower down the market segment order, taking its own sweet time, perhaps this is a clever strategy to make sure existing products are cleared off shelves to make room for new ones, since NVIDIA manufactures in large volumes. Today's release of the GeForce GT 430 graphics processor marks NVIDIA's entry of the Fermi architecture intro the lower-mainstream segment, the segment that holds graphics cards that are geared for GPU-enhanced everyday computing, such as smooth and clear playback of HD video, enhancing the user interface, and GPU-accelerated applications. As far as gaming goes, every game (including DirectX 11 ones) are intended to be playable at lower-resolutions (such as 1280x1024, or 1280x720) by the GeForce GT 430.

Architecture

GeForce GT 430 is based on a brand new chip by NVIDIA, codenamed GF108. The chip is built on the 40 nanometer manufacturing process, continues to be a downscaled design of the Fermi architecture. Key components are arranged in the same order and hierarchy as on every other Fermi GPU. There is one graphics processor cluster (GPC), it shares a common raster engine between two streaming multiprocessors (SMs), each SM shares an instruction cache, a dispatch, and a Polymorph Engine (that carries out several geometry-related functions), between 48 CUDA cores, that are the simplest SIMD units.
GeForce GT430 has 96 CUDA cores on the GPU, 4 raster operations processors (ROPs), and 16 texture memory units. The GPU uses a 128-bit wide memory interface. On this particular SKU there's 1 GB of DDR3 memory.

ZOTAC's GeForce GT 430 uses a custom PCB and thermal solution, while sticking to the performance specifications of the NVIDIA reference design.