原文链接:https://www.cnblogs.com/wmr95/articles/8846749.html
序号名称值解释
1Detected 1 CUDA Capable device(s)1检测到1个可用的NVIDIA显卡设备
2Device 0: "GeForce 930M"GeForce 930M当前显卡型号为" GeForce 930M "
3CUDA Driver Version / Runtime Version7.5/7.5CUDA驱动版本
4CUDA Capability Major/Minor version number5.0CUDA设备支持的计算架构版本,即计算能力,该值越大越好
5Total amount of global memory4096MbytesGlobal memory全局存储器的大小。使用CUDA RUNTIME API调用函数cudaMalloc后,会消耗GPU设备上的存储空间,合理分配和释放空间避免程序出现crash
6(3) Multiprocessors, (128) CUDA Cores/MP384 CUDA Cores3个流多处理器(即SM),每个多处理器中包含128个流处理器,共384个CUDA核
7GPU Max Clock rate941 MHzGPU最大频率
8Memory Clock rate900 MHz显存的频率
9Memory Bus Width64-bit
10L2 Cache Size1048576 bytes
11Maximum Texture Dimension Size (x, y, z)1D=(65535)
2D=(65535, 65535)
3D=(4096,4096,4096)
12Maximum Layered 1D Texture Size, (num) layers1D=(16384),2048 layers
13Maximum Layered 2D Texture Size, (num) layers2D=(16384,16384), 2048 layers
14Total amount of constant memory65535 bytes常量存储器的大小
15Total amount of shared memory per block49152 bytes共享存储器的大小,共享存储器速度比全局存储器快;多处理器上的所有线程块可以同时共享这些存储器
16Total number of registers available per block65535
17Warp Size32Warp,线程束,是SM运行的最基本单位,一个线程束含有32个线程
18Maximum number of threads per multiprocessor2048一个SM中最多有2048个线程,即一个SM中可以有2048/32=64个线程束Warp
19Maximum number of threads per block1024一个线程块最多可用的线程数目
20Max dimension size of a thread block (x, y, z)(1024,1024,64)ThreadIdx.x<=1024,
ThreadIdx.y<=1024,
ThreadIdx.z<=64
Block内三维中各维度的最大值
21Max dimension size of a grid size (x, y, z)(2147483647,65535,65535)Grid内三维中各维度的最大值
22Maximum memory Pitch2147483647 bytes显存访问时对齐时的pitch的最大值
23Texture alignment512 bytes纹理单元访问时对其参数的最大值
24Concurrent copy and kernel executionYes with 1 copy engine(s)
25Run time limit on kernelsYes
26Integrated GPU sharing Host MemoryNo
27Support host page-locked memory mappingYes
28Alignment requirement for SurfacesYes
29Device has ECC supportDisabled
30其他











网友评论