amazon GeForce GT 640 reviews
Preliminary on the card
GK107 chip card or GT 640, using the Kepler architecture (28nm) latest NVIDIA. Compared with the previous Fermi architecture there, characterized by the Kepler graphics core (SP) is much higher clock times and also marginally better. But this does not mean actual performance proportional to the number of workers. If you remember, the GTX 680 the GTX 580 SP 3 times (1536 vs. 512 SP) and also much higher clock speed (772 MHz vs. 1006) but the 28nm card only 1.5 times stronger 40nm seniors . Therefore, do not assume that the SP 384, the GT 640 can stand on a par with the GTX 560 Ti.
Issues outside the power of the GT 640 GK107 chip itself, there is the “contribution” of the DDR3 memory chips. One of the common features of the graphics card is popular segment them or be equipped with low-speed memory (DDR2 or DDR3). Of course there are still versions use a higher speed memory (GDDR5), but obviously the price is also higher. Since DDR3 easier to manufacture and much cheaper cost and GDDR5. This time, the GT 640 encountered the same situation as the retail version is mainly used DDR3 (GDDR5 models on the market only OEM).
But that does not mean all the cards are less used DDR3. The more important is whether the card itself has effectively exploited the memory bandwidth. Close to the GT 640 is GT 440 DDR3 version, with roughly similar memory bandwidth (28.5 vs. 28.8 GB / s). Based on the theory that the NVIDIA® information provided, the amount of memory bandwidth required for the process on the GT 640 stand the lowest in all. Even the ratio GT 440 has bytes per process nearly highest.
A final detailed you also note: GT 640 does not GPU Boost feature (similar to Turbo Boost) as the GTX 670, 680 & 690 same Kepler architecture. Core clock and memory of this card respectively 900 & 1782 MHz, and … that’s all!
Coming to the popular model, the NSX card details often have a “cheat”: VRAM capacity is much higher than the other models. With GT 640, NVIDIA configuration “standard” 2 GB, and the majority of firms are going under this parameter. But you notice that the 2 GB but only DDR3, it can not be more 1 GB GDDR5 . Of course not to say that the amount of VRAM is not important, but with affordable models, few users equipped with screen resolution greater than 1680 x 1050 (or 1600 x 900). Unless gaming on a higher resolution, you have to care about the amount of VRAM (of course the more powerful GPU).
Also, if you notice, AMD just released another update for the HD 7750 with up to 900 MHz clock (800 MHz old version). The version that we reviewed in the post today is the old card. HD 7750 models that are sold with GIGABYTE 880 MHz clock.
where can you get a GeForce GT 640 online
General computing capacity
Capacity game is not the only consideration when buying graphics cards. You can use the GPU for other computing tasks if it supports CUDA, or Direct Stream Compute, OpenCL … With Kepler Architecture (version does not have the dual processor cores real number), the benefits you receive from them even lower than Fermi.
Tessellation is a concern with the capacity of the DX 11 card, as this is the main difference between this API set with DX 10 and the version before. NVIDIA has been rising at the Fermi architecture with breakthrough AMD and launch GCN architecture. Kepler architecture is generally not very strong as the GPU of GT 640 to the 1.3 billion trans while GPU of GTS 450 only 1.17 billion trans.