Unable to install CUDA 10.0 [Ubuntu 18.04 LTS - Dell XPS 15 9570]

Accelerated Computing CUDA CUDA Setup and Installation
#1

Hi, I am trying to install CUDA 10.0 using the .deb file following this guide https://developer.download.nvidia.com/compute/cuda/10.0/Prod/docs/sidebar/CUDA_Installation_Guide_Linux.pdf

When I try to execute this command

sudo /usr/bin/nvidia-persistenced --verbose

I got the error

nvidia-persistenced failed to initialize. Check syslog for more details.

These are the relevant syslogs

Mar  9 11:51:34 amarchin-XPS-15-9570 nvidia-persistenced: Started (5899)
Mar  9 11:51:34 amarchin-XPS-15-9570 kernel: [ 2000.328437] pci 0000:01:00.0: Refused to change power state, currently in D3
Mar  9 11:51:34 amarchin-XPS-15-9570 nvidia-persistenced: Failed to query NVIDIA devices. Please ensure that the NVIDIA device files (/dev/nvidia*) exist, and that user 0 has read and write permissions for those files.
Mar  9 11:51:34 amarchin-XPS-15-9570 nvidia-persistenced: PID file unlocked.
Mar  9 11:51:34 amarchin-XPS-15-9570 nvidia-persistenced: PID file closed.
Mar  9 11:51:34 amarchin-XPS-15-9570 nvidia-persistenced: Shutdown (5899)

Do you have any suggestion?

#2

Your laptop has the GPU in a sleep state. You won’t be able to install the persistence daemon that way.

https://docs.microsoft.com/en-us/windows-hardware/drivers/kernel/device-sleeping-states#d3

I would either skip this step (enabling of persistence daemon) or else inspect your laptop system BIOS setup to see if there is a way to force the GPU on.

#3

Hi Robert, thank you. So I skipped this and I am dealing with “7.2.3.2. Compiling the Examples”.
When I am in the folder ~/NVIDIA_CUDA-10.0_Samples and try to compile I get

➜  NVIDIA_CUDA-10.0_Samples make
make[1]: Entering directory '/home/amarchin/NVIDIA_CUDA-10.0_Samples/0_Simple/cudaTensorCoreGemm'
make[1]: Nothing to be done for 'all'.
make[1]: Leaving directory '/home/amarchin/NVIDIA_CUDA-10.0_Samples/0_Simple/cudaTensorCoreGemm'
make[1]: Entering directory '/home/amarchin/NVIDIA_CUDA-10.0_Samples/0_Simple/simpleCudaGraphs'
make[1]: Nothing to be done for 'all'.
make[1]: Leaving directory '/home/amarchin/NVIDIA_CUDA-10.0_Samples/0_Simple/simpleCudaGraphs'
make[1]: Entering directory '/home/amarchin/NVIDIA_CUDA-10.0_Samples/0_Simple/vectorAdd'
make[1]: Nothing to be done for 'all'.
make[1]: Leaving directory '/home/amarchin/NVIDIA_CUDA-10.0_Samples/0_Simple/vectorAdd'
make[1]: Entering directory '/home/amarchin/NVIDIA_CUDA-10.0_Samples/0_Simple/fp16ScalarProduct'
make[1]: Nothing to be done for 'all'.
make[1]: Leaving directory '/home/amarchin/NVIDIA_CUDA-10.0_Samples/0_Simple/fp16ScalarProduct'
make[1]: Entering directory '/home/amarchin/NVIDIA_CUDA-10.0_Samples/0_Simple/simpleStreams'
make[1]: Nothing to be done for 'all'.
make[1]: Leaving directory '/home/amarchin/NVIDIA_CUDA-10.0_Samples/0_Simple/simpleStreams'
make[1]: Entering directory '/home/amarchin/NVIDIA_CUDA-10.0_Samples/0_Simple/simpleTemplates_nvrtc'
make[1]: Nothing to be done for 'all'.
make[1]: Leaving directory '/home/amarchin/NVIDIA_CUDA-10.0_Samples/0_Simple/simpleTemplates_nvrtc'
make[1]: Entering directory '/home/amarchin/NVIDIA_CUDA-10.0_Samples/0_Simple/simpleAtomicIntrinsics_nvrtc'
make[1]: Nothing to be done for 'all'.
make[1]: Leaving directory '/home/amarchin/NVIDIA_CUDA-10.0_Samples/0_Simple/simpleAtomicIntrinsics_nvrtc'
make[1]: Entering directory '/home/amarchin/NVIDIA_CUDA-10.0_Samples/0_Simple/cppOverload'
make[1]: Nothing to be done for 'all'.
make[1]: Leaving directory '/home/amarchin/NVIDIA_CUDA-10.0_Samples/0_Simple/cppOverload'
make[1]: Entering directory '/home/amarchin/NVIDIA_CUDA-10.0_Samples/0_Simple/simpleOccupancy'
make[1]: Nothing to be done for 'all'.
make[1]: Leaving directory '/home/amarchin/NVIDIA_CUDA-10.0_Samples/0_Simple/simpleOccupancy'
make[1]: Entering directory '/home/amarchin/NVIDIA_CUDA-10.0_Samples/0_Simple/simpleAssert_nvrtc'
make[1]: Nothing to be done for 'all'.
make[1]: Leaving directory '/home/amarchin/NVIDIA_CUDA-10.0_Samples/0_Simple/simpleAssert_nvrtc'
make[1]: Entering directory '/home/amarchin/NVIDIA_CUDA-10.0_Samples/0_Simple/asyncAPI'
make[1]: Nothing to be done for 'all'.
make[1]: Leaving directory '/home/amarchin/NVIDIA_CUDA-10.0_Samples/0_Simple/asyncAPI'
make[1]: Entering directory '/home/amarchin/NVIDIA_CUDA-10.0_Samples/0_Simple/simpleLayeredTexture'
make[1]: Nothing to be done for 'all'.
make[1]: Leaving directory '/home/amarchin/NVIDIA_CUDA-10.0_Samples/0_Simple/simpleLayeredTexture'
make[1]: Entering directory '/home/amarchin/NVIDIA_CUDA-10.0_Samples/0_Simple/inlinePTX'
make[1]: Nothing to be done for 'all'.
make[1]: Leaving directory '/home/amarchin/NVIDIA_CUDA-10.0_Samples/0_Simple/inlinePTX'
make[1]: Entering directory '/home/amarchin/NVIDIA_CUDA-10.0_Samples/0_Simple/simpleCooperativeGroups'
make[1]: Nothing to be done for 'all'.
make[1]: Leaving directory '/home/amarchin/NVIDIA_CUDA-10.0_Samples/0_Simple/simpleCooperativeGroups'
make[1]: Entering directory '/home/amarchin/NVIDIA_CUDA-10.0_Samples/0_Simple/simpleSeparateCompilation'
make[1]: Nothing to be done for 'all'.
make[1]: Leaving directory '/home/amarchin/NVIDIA_CUDA-10.0_Samples/0_Simple/simpleSeparateCompilation'
make[1]: Entering directory '/home/amarchin/NVIDIA_CUDA-10.0_Samples/0_Simple/matrixMul_nvrtc'
cp ""/usr/local/cuda-10.0"/include/cooperative_groups.h" .
cp: cannot stat '/usr/local/cuda-10.0/include/cooperative_groups.h': No such file or directory
Makefile:323: recipe for target 'build' failed
make[1]: *** [build] Error 1
make[1]: Leaving directory '/home/amarchin/NVIDIA_CUDA-10.0_Samples/0_Simple/matrixMul_nvrtc'
Makefile:51: recipe for target '0_Simple/matrixMul_nvrtc/Makefile.ph_build' failed
make: *** [0_Simple/matrixMul_nvrtc/Makefile.ph_build] Error 2

It seems that in the folder /usr/local/cuda-10.0/include/ there is only the file cudnn.h.

#4

Then there is something wrong with your install.