You can set the device order for CUDA environment in your shell to follow the bus ID instead of the default of fastest card. Requires CUDA 7 and up.
export CUDA_DEVICE_ORDER=PCI_BUS_ID
More Related Contents:
- Unspecified launch failure on Memcpy
- Different CUDA versions shown by nvcc and NVIDIA-smi
- Thrust inside user written kernels
- Modifying registry to increase GPU timeout, windows 7
- Understanding CUDA grid dimensions, block dimensions and threads organization (simple explanation) [closed]
- What is the purpose of using multiple “arch” flags in Nvidia’s NVCC compiler?
- How are 2D / 3D CUDA blocks divided into warps?
- What can I do against ‘CUDA driver version is insufficient for CUDA runtime version’?
- What is a bank conflict? (Doing Cuda/OpenCL programming)
- nvidia-smi Volatile GPU-Utilization explanation?
- Can anyone provide sample code demonstrating the use of 16 bit floating point in cuda?
- Any particular function to initialize GPU other than the first cudaMalloc call?
- CUDA and nvcc: using the preprocessor to choose between float or double
- Polymorphism and derived classes in CUDA / CUDA Thrust
- What kind of variables consume registers in CUDA?
- CUDA: How many concurrent threads in total?
- CUDA apps time out & fail after several seconds – how to work around this?
- How do I select which GPU to run a job on?
- multi-GPU basic usage
- How to use 2D Arrays in CUDA?
- device function pointers
- How can I make tensorflow run on a GPU with capability 2.x?
- cudaMemset() – does it set bytes or integers?
- How to implement device side CUDA virtual functions?
- How to create a CUDA context?
- Why does my CUDA kernel crash (unspecified launch failure) with a different dataset size?
- CUDA allocation alignment is 256 bytes – seriously?
- CUDA function pointers
- Which Compute Capability is supported by which CUDA versions?
- Using an array of device function pointers