GPU coder error: Invalid ComputeCapability value 7.5

I tried to verify GPU code generation environment with example code below, and got the following error. Does ComputeCapability 7.5 too high to compute?

gpuEnvObj = coder.gpuEnvConfig;
gpuEnvObj.GpuId = 1;
gpuEnvObj.BasicCodegen = 1;
gpuEnvObj.BasicCodeexec = 1;
results = coder.checkGpuInstall(gpuEnvObj)

Error using coder.checkGpuInstall (line 32)

Error setting property 'ComputeCapability' of class 'GpuConfig': Invalid value '7.5'.

Allowed values are:

3.2, 3.5, 3.7, 5.0, 5.2, 5.3, 6.0, 6.1, 6.2, 7.0, 7.1, 7.2

Answer 1

[Walter Roberson]

You have a Turing class NVIDIA GPU, but your driver supports only up to Volta.

It looks to me as if you are probably using a 300 series driver such as 384.111

You can update the driver on your system to any 400 series driver. You might as well use the newest available driver for your system, as the newest supports all architectures that your current driver supports, and more.

[shuang zhou]

I updated to Game Ready Driver Version 452.06, and got the same result.

gpuEnvObj = coder.gpuEnvConfig;

gpuEnvObj.GpuId = 0;

gpuEnvObj.BasicCodegen = 1;

gpuEnvObj.BasicCodeexec = 1;

results = coder.checkGpuInstall(gpuEnvObj)

Compatible GPU : PASSED

CUDA Environment : PASSED

Runtime : PASSED

cuFFT : PASSED

cuSOLVER : PASSED

cuBLAS : PASSED

Error using coder.checkGpuInstall (line 32)

Error setting property 'ComputeCapability' of class 'GpuConfig': Invalid value '7.5'.

Allowed values are:

3.2, 3.5, 3.7, 5.0, 5.2, 5.3, 6.0, 6.1, 6.2, 7.0, 7.1, 7.2

[shuang zhou]

CUDADevice with properties:

Name: 'GeForce RTX 2080'

Index: 1

ComputeCapability: '7.5'

SupportsDouble: 1

DriverVersion: 11

ToolkitVersion: 10

MaxThreadsPerBlock: 1024

MaxShmemPerBlock: 49152

MaxThreadBlockSize: [1024 1024 64]

MaxGridSize: [2.1475e+09 65535 65535]

SIMDWidth: 32

TotalMemory: 8.5899e+09

AvailableMemory: 7.3315e+09

MultiprocessorCount: 46

ClockRateKHz: 1800000

ComputeMode: 'Default'

GPUOverlapsTransfers: 1

KernelExecutionTimeout: 1

CanMapHostMemory: 1

DeviceSupported: 1

DeviceSelected: 1

[Walter Roberson]

That "should" work; https://www.mathworks.com/help/parallel-computing/gpu-support-by-release.html indicates that capacity 7.5 is supported for your release.

But just to check: did you reboot after you installed the newest driver> You need to have rebooted after installing the driver.

[shuang zhou]

My current matlab version is 2019a, I solved this problem by update matlab to 2020b prerelease .

Then add NVIDIA_CUDNN and NVIDIA_TENSORRT to system environment variables list

>> getenv('NVIDIA_CUDNN')

'C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v9.2'

>> getenv('NVIDIA_TENSORRT')

'C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v9.2\TensorRT-7.1.3.4'

[shuang zhou]

MATLAB 2019A

MATLAB 2020B