[CUDA Toolkit v10.1] cudaGetDeviceProperties returnes 999 (cudaErrorUnknown)

The cudaGetDeviceProperties() API returns 0 or 999 with CUDA Toolkit Samples (1_Utilities\deviceQuery).
And the probability of returning 999 is high, when stepping in debug mode.

I have tried other versions of the CUDA Toolkit, but only v10.1 has the problem, so I guess that something has changed between v10.0 and v10.1.

Environment:

  • OS : Windows 10 (build 17763.615)
  • IDE: Visual Studio 2017 Update 9 (version 15.9.14)
  • GPU: GeForce GT 730M

CUDA Toolkit versions:

  • v9.2 works fine.
  • v10.0 works fine.
  • v10.1 has problem.

I’m waiting for any help.
Regards.

I’m sorry I can’t help because it all works just fine for me. I use W10 with VS2017 but the lowest-end GPU I have is a 970M.

Here’s a screenshot of my little test program in action.

Thank you for reply, ryork. I know v10.1 works fine in your environment.

The result of deviceQuery.exe in case of succeeded with Toolkit v10.0 in my environment is follows.

deviceQuery.exe Starting...

 CUDA Device Query (Runtime API) version (CUDART static linking)

Detected 1 CUDA Capable device(s)

Device 0: "GeForce GT 730M"
  CUDA Driver Version / Runtime Version          10.0 / 10.0
  CUDA Capability Major/Minor version number:    3.5
  Total amount of global memory:                 1024 MBytes (1073741824 bytes)
  ( 2) Multiprocessors, (192) CUDA Cores/MP:     384 CUDA Cores
  GPU Max Clock rate:                            758 MHz (0.76 GHz)
  Memory Clock rate:                             1001 Mhz
  Memory Bus Width:                              64-bit
  L2 Cache Size:                                 524288 bytes
  Maximum Texture Dimension Size (x,y,z)         1D=(65536), 2D=(65536, 65536), 3D=(4096, 4096, 4096)
  Maximum Layered 1D Texture Size, (num) layers  1D=(16384), 2048 layers
  Maximum Layered 2D Texture Size, (num) layers  2D=(16384, 16384), 2048 layers
  Total amount of constant memory:               65536 bytes
  Total amount of shared memory per block:       49152 bytes
  Total number of registers available per block: 65536
  Warp size:                                     32
  Maximum number of threads per multiprocessor:  2048
  Maximum number of threads per block:           1024
  Max dimension size of a thread block (x,y,z): (1024, 1024, 64)
  Max dimension size of a grid size    (x,y,z): (2147483647, 65535, 65535)
  Maximum memory pitch:                          2147483647 bytes
  Texture alignment:                             512 bytes
  Concurrent copy and kernel execution:          Yes with 1 copy engine(s)
  Run time limit on kernels:                     Yes
  Integrated GPU sharing Host Memory:            No
  Support host page-locked memory mapping:       Yes
  Alignment requirement for Surfaces:            Yes
  Device has ECC support:                        Disabled
  CUDA Device Driver Mode (TCC or WDDM):         WDDM (Windows Display Driver Model)
  Device supports Unified Addressing (UVA):      Yes
  Device supports Compute Preemption:            No
  Supports Cooperative Kernel Launch:            No
  Supports MultiDevice Co-op Kernel Launch:      No
  Device PCI Domain ID / Bus ID / location ID:   0 / 1 / 0
  Compute Mode:
     < Default (multiple host threads can use ::cudaSetDevice() with device simultaneously) >

deviceQuery, CUDA Driver = CUDART, CUDA Driver Version = 10.0, CUDA Runtime Version = 10.0, NumDevs = 1
Result = PASS

By the way, I peformed some experiments with the combination of runtime library and driver, and I think the problem might be in driver version 425.25 or later from following result.

  • cudart64_100.dll (v10.0) x driver 411.31 (v10.0) -> cudaGetDeviceProperty() returns 0.
  • cudart64_100.dll (v10.0) x driver 425.25 (v10.1) -> cudaGetDeviceProperty() returns 0 or 30 (defined as cudaErrorUnknown in v10.0)
  • cudart64_101.dll (v10.1) x driver 425.25 (v10.1) -> cudaGetDeviceProperty() returns 0 or 999 (defiend as cudaErrorUnknown in v10.1)
  • cudart64_101.dll (v10.1) x driver 425.31 (Game Ready Driver) -> cudaGetDeviceProperty() returns 0 or 999 (defiend as cudaErrorUnknown in v10.1)

Please let me know if anyone have additional information or notices.
Regards,