GeForce GT 430
The insides of the GF108 chip are two clusters of shader units version GF104/106, with 48 shader processors and 8 TMUs each. Although, contrary to fears, NVIDIA did not reduce the performance of the geometry, retaining the Polymorph Engine block per cluster, the ROP blocks were “castrated” - instead of the “8 ROP per 64-bit bus” principle, only 108 ROP blocks were left in the GF4. This means that the already modest theoretical performance of the solution in real applications will be absolutely terrible.
The reference design has a 1U-oriented design, DVI, HDMI and VGA ports, an active cooling system, but manufacturers of final cards will certainly offer their own solutions, including those with passive cooling. The card's power consumption is very modest, the card does not require power amplification, and SLI support is considered redundant functionality for the entry level.
Characteristics of NVIDIA GeForce GT 430
NameGeForce GT 430CoreGF108Technical process (µm)0.04Transistors (million)?Core frequency700Memory frequency (DDR)900 (1800)Bus and memory typeGDDR3 128-bitBPS (Gb/s)28.8Unified shader units96Unified shader units frequency 1400TMU per conveyor16ROP16Shaders Model5.0Fill Rate ( Mpix/s)11200Fill Rate (Mtex/s)11200DirectX11.0Memory capacity1024InterfacePCI-E 2.0 x16
In general, the new solution is an analogue of the GeForce GT 240 with a new architecture and support for DirectX 11/OpenGL 4/CUDA compute capability 2.1, but a careful look will immediately reveal that the new product differs in a number of parameters, including video memory bandwidth and scene fill rate not for the better. With such a strategy, it will be extremely difficult to beat competing AMD Radeon solutions in the same price range (recommended price: $79).
