nvcc fatal : Unsupported gpu architecture 'compute_13' ??

Please help me. I just got a new device with gtx970m + matlab2015b+ cuda8.0 + windows10 + visual studio 2012/2010. but when i try to compile a .cu file in matlab, it tells me this:
nvcc fatal : Unsupported gpu architecture 'compute_13'
C:\PROGRA~1\MATLAB\R2013A\BIN\MEX.PL: Error: Compile of 'pf_cuda.cu' failed.
matlab can recognize the gpu: when i type: d = gpuDevice it returns:
CUDADevice with properties:
Name: 'GeForce GTX 970M'
Index: 1
ComputeCapability: '5.2'
SupportsDouble: 1
DriverVersion: 8
ToolkitVersion: 5
MaxThreadsPerBlock: 1024
MaxShmemPerBlock: 49152
MaxThreadBlockSize: [1024 1024 64]
MaxGridSize: [2.1475e+09 65535 65535]
SIMDWidth: 32
TotalMemory: 6.4425e+09
FreeMemory: 5.3517e+09
MultiprocessorCount: 10
ClockRateKHz: 1038000
ComputeMode: 'Default'
GPUOverlapsTransfers: 1
KernelExecutionTimeout: 1
CanMapHostMemory: 1
DeviceSupported: 1
DeviceSelected: 1
Can anybody help me? Many thanks.

답변 (1개)

Edric Ellis
Edric Ellis 2017년 2월 27일
편집: Edric Ellis 2017년 2월 27일

1 개 추천

To create GPU MEX files, you need to ensure that you have installed the version of the CUDA toolkit corresponding to the one used by MATLAB. For R2015b, that's CUDA 7.0. Are you sure you're using R2015b though - the MEX path mentions R2013a, and the gpuDevice displayed output indicates an older version of the CUDA toolkit than shipped with R2015b.

카테고리

도움말 센터File Exchange에서 Get Started with GPU Coder에 대해 자세히 알아보기

질문:

2017년 2월 25일

댓글:

2017년 7월 24일

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!

Translated by