Overview
GeForce GT 640 is a Middle-Class Graphics Card based on the first revision of the Kepler Architecture.

Architecture
The Kepler Architecture was NVIDIA's big step to power efficiency. Each Stream Multiprocessor (SMX) now hosts 192 Shader Processing Units - against the 48 of older Fermi Architecture, and has been redesigned being now clocked at the same speed of the Central Unit. This means they are more energy efficient and will consequently lead to cooler operating temperatures. However, it also means they are weaker. It can be said that one Fermi SMX is as fast as 2 Kepler SMXs.
Additionally, and not available in all GPUs, Kepler also introduced the Boost Clock Feature. The Boost Clock is an even higher Clock Speed activated when in gaming mode and becomes the effective speed of the GPU.

GPU
It equips a GPU Codenamed GK107-300-A2 which has 2 Stream Multiprocessor activated and thus offers 384 Shader Processing Units, 32 TMUs and 16 ROPs. The Central Unit is clocked at 902MHz.

Memory
The GPU accesses a 2GB frame buffer of DDR3, through a 128-bit memory interface. The size of the frame buffer is exaggerated and in no way benefits the GPU. The Memory Clock Operates at 891MHz.

Features
DirectX 11.0 Support (11.0 Hardware Default) and support for Optimus, CUDA, OpenCL, DirectCompute, 3D Vision Surround, PhysX, Realtime Raytracing and other technologies

Power Consumption
With a rated board TDP of 65W, it requires at least a 350W PSU and it relies entirely on the PCI Slot for power, meaning no extra connectors are required.

Performance
Gaming benchmarks put its performance on average with Radeon HD 6670.

System Suggestions
We recommend a Modest Processor (Intel Core i3) and 8GB of RAM for a system with GeForce GT 640.