NVIDIA CUDA 3.0 and OpenCL SDKs with Fermi Support Available

我酷故我在

NVIDIA has released a new version of its GPU Computing SDK. This version supports Fermi architecture and will allow GPU computing developers to prepare their code for Fermi-based graphics cards. GPU Computing SDK is made up of CUDA 3.0 Toolkit as well as the OpenCL SDK.
Main new features:

• Support for the new Fermi architecture (native 64-bit GPU support, ECC reporting, …)
• A new unified interoperability API for Direct3D and OpenGL
• New CUDA Memory Checker
• Support for all the OpenCL features in the latest R195 production driver package (double precision, graphics interoperability with OpenCL, Direc3D9, Direct3D10, and Direct3D11)
• Support for the latest OpenCL spec revision 1.0.48 and latest official Khronos OpenCL headers as of 2010-02-17
• Tesla Compute Cluster (TCC)

Downloads:

• CUDA Toolkit 3.0 Windows 32-bit
• CUDA Toolkit 3.0 Windows 64-bit
• GPU Computing CUDA SDK Windows 64-bit
• GPU Computing CUDA SDK Windows 32-bit

More information (and other downloads) about the NVIDIA’s GPU Computing SDK can be found HERE and HERE.

xiao069532

能否翻译下。。。。。。。。。。

xaao

路过，拿3分活跃指数，对CUDA本身无视之

sciwander

软件早有了，卡到底啥时候有泥

我酷故我在

NVIDIA今天正式发布了全新一代通用计算开发工具包“CUDA 3.0 Toolkit”，特别加入了对Fermi架构新一代显卡和C++编程语言的支持，并有其他诸多新特性。
CUDA 3.0支持基于Fermi架构的GeForce GTX 400系列、Quadro、Tesla 20系列显卡，包括：原生64位GPU、多重复制引擎、ECC报告、并存内核执行、cuda-gdb支持Fermi硬件调试、Visual Profiler CUDA C/OpenCL Fermi硬件分析。
另外值得关注的是，CUDA 3.0首次支持C++类继承和模板继承，从而提高编程灵活性，同时CUDA C/C++内核现在以标准ELF格式进行编译，还增加了一个新的Direct3D、OpenGL统一协作API，支持OpenGL纹理和Direct3D 11。其他众多开发新特性不再赘述。
NVIDIA CUDA 3.0支持Windows XP、Windows Vista、Windows 7、Fedora 10、RedHat Enterprise Linux 5.3/4.8、Ubuntu 9.04、OpenSUSE 11.1、SUSE Linux Enterprise Desktop 11、Mac OS X等众多操作系统的32/64位版本。
CUDA 3.0详细说明和相关资源下载：
http://developer.nvidia.com/object/cuda_3_0_downloads.html

基于Fermi架构的Tesla C2050

jocover

终于知道那几个nv opencl扩展怎么用了

fpq

cuda很有前途，但目前cuda在民用级的产品里还没多少实际意义

seraphdoo

不错，希望今后的软件更多的利用这些运算技术。

晨之晖

说实话目前看CUDA比Phsxy更有前途

jiujiu

路过，拿3分活跃指数