The latest GTX 400 series from Nvidia was a flop, which is beyond question. Although the cards provide a high Tessellationsleistung disappoint under DirectX 11, but otherwise known with little extra power at a high price and above all by their exorbitant power consumption.
Never theless, or because of it, now will the new Nvidia CUDA Toolkit version 3.2 extra on the maps of the Fermi line, resulting in significant speed improvements in the CUDA / PhysX implementations. Thereby increasing the velocity for resizing 3D objects by a factor of 3-7, depending on which chip is used. The matrix multiplications, which are used inter alia in the calculation of data from SETI @ Home and similar clients, generally between 50% and 300% faster. In addition, the acceleration of H.264 video now part of the CUDA toolkit, so that the developer no longer has to grapple with the GPU SDK. One can say that it is mainly intended for game designers for specific optimizations. New is also support for Nvidia Quadro and Tesla cards with 6 GB of video memory, or the debug function on multi-GPU systems.
More over, applying themselves to the Intel compiler, which may on the announced SSE2 could point optimization, which would CUDA accelerate calculation on systems without a Nvidia graphics card significantly - currently this only the x86 instruction unit is used, which of the two CPU forging very neglected treated, since it is SSE far inferior in all its extension forms, both in programming and in performance. The improved integration of Microsoft Visual Studio to help developers to rapidly implement and combine good features from both systems.
Never theless, or because of it, now will the new Nvidia CUDA Toolkit version 3.2 extra on the maps of the Fermi line, resulting in significant speed improvements in the CUDA / PhysX implementations. Thereby increasing the velocity for resizing 3D objects by a factor of 3-7, depending on which chip is used. The matrix multiplications, which are used inter alia in the calculation of data from SETI @ Home and similar clients, generally between 50% and 300% faster. In addition, the acceleration of H.264 video now part of the CUDA toolkit, so that the developer no longer has to grapple with the GPU SDK. One can say that it is mainly intended for game designers for specific optimizations. New is also support for Nvidia Quadro and Tesla cards with 6 GB of video memory, or the debug function on multi-GPU systems.
More over, applying themselves to the Intel compiler, which may on the announced SSE2 could point optimization, which would CUDA accelerate calculation on systems without a Nvidia graphics card significantly - currently this only the x86 instruction unit is used, which of the two CPU forging very neglected treated, since it is SSE far inferior in all its extension forms, both in programming and in performance. The improved integration of Microsoft Visual Studio to help developers to rapidly implement and combine good features from both systems.