incompatibility problem

hi all,

my system conf. is:

ubuntu10.10 32bit
nvidia driver version 260.19.06
gcc version 4.4.5
gpu version GeForce GTX 460
cuda toolkit version 4.0
gpu computing sdk version 4.0.17

And when I try to compile SDK samples I get:

bla bla bla
make[1]: Entering directory /home/buse/NVIDIA_GPU_Computing_SDK/C/src/recursiveGaussian' /usr/bin/ld: cannot find -lcudart /usr/bin/ld: cannot find -lcudart collect2: ld returned 1 exit status make[1]: *** [../../bin/linux/release/recursiveGaussian] Error 1 make[1]: Leaving directory /home/buse/NVIDIA_GPU_Computing_SDK/C/src/recursiveGaussian’
make: *** [src/recursiveGaussian/Makefile.ph_build] Error 2

and when I try to run my own source I get:

buse@ubuntu:~/gpgpu_itu/lab$ nvcc warning: variable “numBlocks” was declared but never referenced warning: variable “numBlocks” was declared but never referenced

buse@ubuntu:~/gpgpu_itu/lab$ ./a.out
Cuda error: kernel execution: CUDA driver version is insufficient for CUDA runtime version.

seems to me as if I have a compatibility prob. don^t know how to solve.
Now doesn’t gcc 4.5 is supported by cuda 4.0?
how can I get an older version of cuda?

CUDA 4.0 requires a 270 driver.

Go to and follow the link to the toolkit download.

You can install an old toolkit on a new driver but not a new toolkit on a old driver

thank you for the post mfatica, I did as you said. now I have a fully functional cuda running on my system (cuda3.2) :happy: