Cuda 4.0 on Ubuntu 10.04 LTS

Why can’t we install cuda 4.0 on Ubuntu 10.04 Long term support?
I ask because our server admin is unwilling to use the OS without LTS, i.e. 10.10.

I also would like to install CUDA 4.0 onto Ubuntu 10.04. How do I do this?

try using the (sudo -i) command.

the problem is that cuda 4.0 is NOT compatible with 10.04, not that it can’t be physically installed. So the question is why does cuda toolkit 4.0 requires ubuntu 10.10 and won’t work on 10.04 ?

I don’t understand why you think CUDA 4.0 is incompatible with 10.04 LTS. I have been running both release candidates and the final of 4.0 on 64 bit 10.04 LTS for months with no problems. What doesn’t work for you?

avidday:

I’ve been looking for you for two weeks! Are these the steps you followed to install CUDA 4.0 onto Ubuntu 10.04:

  1. Install ubuntu 10.04

  2. Visit http://developer.nvidia.com/cuda-toolkit-40

  3. Download “CUDA Toolkit for Ubuntu Linux 10.10” (eventhough you have Ubuntu 10.04)

  4. Execute the downloaded file with no complaints from the installer about your version of Ubuntu, GCC, etc.

Thank you,

Chris

CUDA 4.0 and Ubuntu 10.04 LTS also works for me. Below is my installation way. Hope to be assistant.

1. Install driver
CUDA driver is broken after linux kernel update, need to reinstall driver after the update.

a) Stop X
Logout and go to 1st console if you are using X11.
(Switch to other consoles by pressing Ctrl-Alt-F1, etc.)

$ sudo stop gdm

b) Install cuda driver

$ sudo sh devdriver_4.0_linux_64_270.41.19.run

Obviously, I’m using 64-bit Ubuntu. Modify this line to your settings.

From 270.35, cuda driver provides nvidia-xconfig utility to automatically update X configuration file so that driver will be used when restarting X. I alway refuse to use it since my way works well. If anyone tried it, let me know the details.

This program will uninstall old driver first (if you have). Something about pre-install script error will occur, choose ‘Continue’. And I also choose to install 32-big compatibility OpenGL libraries.

c) (Optional) Install boot script
(If script already installed previously, cancel this step.)
(See below for the content of script ‘loadCudaModule’.)

$ sudo cp loadCudaModule /etc/init.d

$ sudo chmod +x /etc/init.d/loadCudaModule

$ cd /etc/init.d

/etc/init.d$ sudo update-rc.d loadCudaModule start 51 S .

d) Reboot computer

e) Verify driver

$ sudo nvidia-xconfig -query-gpu-info

$ cat /proc/driver/nvidia/version

2. Install toolkit

a) Uninstall old toolkit (if you have)

$ sudo rm -fr /usr/local/cuda

b) Install new toolkit

$ sudo sh cudatoolkit_4.0.17_linux_64_ubuntu10.10.run

c) Verify toolkit

$ nvcc -V

3. Install SDK

a) Uninstall old SDK (if you have)

$ rm -fr ~/NVIDIA_GPU_Computing_SDK

b) Install new SDK

$ sh gpucomputingsdk_4.0.17_linux.run

~/NVIDIA_GPU_Computing_SDK/C$ make

c) Verify SDK

$ ~/NVIDIA_GPU_Computing_SDK/C/bin/linux/release/deviceQuery

# The content of File ‘loadCudaModule’
#!/bin/bash

modprobe nvidia

if [ “$?” -eq 0 ]; then

Count the number of NVIDIA controllers found.

N3D=lspci | grep -i NVIDIA | grep "3D controller" | wc -l
NVGA=lspci | grep -i NVIDIA | grep "VGA compatible controller" | wc -l

N=expr $N3D + $NVGA - 1
for i in seq 0 $N; do
mknod -m 666 /dev/nvidia$i c 195 $i;
done

mknod -m 666 /dev/nvidiactl c 195 255

else
exit 1
fi

mainly because the release notes of 4.0 all said it didn’t support 10.04 anymore.

I went back and tried to install and link with 4.0 runtime and seems to work so far.

thanks

I had been using CUDA 4.0 for a long time with ubuntu 10.04 LTS and never experienced any problems. But after extending the functionality of my program and using e.g the cufft library strange errors started occuring. Took me a while, but the problem seemed to be some sort of compatibility issue between CUDA 4.0 and Ubuntu 10.04. Rolled back drivers to 3.2 and everything worked as it was supposed to. Anyways; if strange things start happening it might be something to keep in the back of your head :)

Hello,

I have problems with the installations. I had the 3.2 version installed, then I install the 4.0 version. I am not able to run most of the examples.

A few of the errors:

C/bin/linux/release/./particles Starting…

grid: 64 x 64 x 64 = 262144 cells

particles: 16384

particleSystem.cu(88) : cudaSafeCall() Runtime API error 3: initialization error.

C/bin/linux/release/./deviceQuery

[deviceQuery] starting…

C/bin/linux/release/./deviceQuery Starting…

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

cudaGetDeviceCount returned 3

-> initialization error

[deviceQuery] test results…

FAILED

Press ENTER to exit…

./oceanFFT

[CUDA FFT Ocean Simulation]

Left mouse button - rotate

Middle mouse button - pan

Right mouse button - zoom

‘w’ key - toggle wireframe

[CUDA FFT Ocean Simulation]

CUFFT error in file ‘oceanFFT.cpp’ in line 324.

C/bin/linux/release/./bandwidthTest

[bandwidthTest] starting…

C/bin/linux/release/./bandwidthTest Starting…

Running on…

cudaGetDeviceProperties returned 3

-> initialization error

[bandwidthTest] test results…

FAILED

Press ENTER to exit…

C/bin/linux/release/./deviceQuery

[deviceQuery] starting…

C/bin/linux/release/./deviceQuery Starting…

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

cudaGetDeviceCount returned 3

-> initialization error

[deviceQuery] test results…

FAILED

Press ENTER to exit…

What have I done wrong the example that have the Drv at the end run without problem.

If you passed all those driver and toolkit verifications, you may want to try newer version of driver, as stated in here, or google more “cudaGetDeviceCount returned 3”.

Thank you. I will try it. For some reason I thought the 270 version which I have now installed was the latest,