SCIENTIFIC-LINUX-USERS Archives

April 2017

SCIENTIFIC-LINUX-USERS@LISTSERV.FNAL.GOV

Options: Use Monospaced Font
Show HTML Part by Default
Show All Mail Headers

Message: [<< First] [< Prev] [Next >] [Last >>]
Topic: [<< First] [< Prev] [Next >] [Last >>]
Author: [<< First] [< Prev] [Next >] [Last >>]

Print Reply
Subject:
From:
"Lofgren, Eric" <[log in to unmask]>
Reply To:
Lofgren, Eric
Date:
Tue, 18 Apr 2017 23:30:21 +0000
Content-Type:
multipart/alternative
Parts/Attachments:
text/plain (1264 bytes) , text/html (2320 bytes)
Hey all,



I’m relatively new to Linux generally and SL in particular. I’m running SL7, and have run aground trying to follow the setup instructions at http://developer.download.nvidia.com/compute/cuda/8.0/secure/Prod2/docs/sidebar/CUDA_Installation_Guide_Linux.pdf?autho=1492553753_fc0d9c1b985f0ba78200964ccf8f1dda&file=CUDA_Installation_Guide_Linux.pdf



I’ve been following the RHEL/CentOS instructions, as my understanding is that SL7 should just be a drop in for those. Things seem to go well until the “Verify the Installation” step at 6.2.2 in the document.



Verifying the Driver Version with $cat /proc/driver/nvidia/version returns ‘No such file or directory’



The examples compile fine in 6.2.2.3, but when you run the deviceQuery binary, rather than finding the Quadro K620 sitting in the machine at the moment, it returns:



./deviceQuery Starting…



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



cudaGetDeviceCount returned 30

-> unknown error

Result = FAIL



That’s where I stopped, as it appears the GPU isn’t talking nicely with CUDA. lspci can see the GPU - is this just a driver issue? If so, is sudo yum install cuda-drivers likely the easiest fix?



Thanks in advance for the help,



Eric


ATOM RSS1 RSS2