CUDA on Alienware M11xR2(optimus) Can't install developer driver

I would like to do some CUDA development on my alienware m11xR2 laptop (the one with the dreaded optimus…i.e. no linux support from nvidia…). Actually, I bought it for work and I NEED to do some CUDA on it…

It has the NVIDIA Geforce 335M which I am pretty sure has CUDA support.

I have Win7-64 bit OS.

I have downloaded and tried to install the latest developer driver, devdriver_3.2_winvista-win7_64_260.61_notebook, but when I try to install it, it says that no supported hardware can be found.

What are my options? Am I trying to install the wrong driver?

Surely this is possible. I mean… I know linux isn’t supported but this is Windows 7…

CD

I would like to do some CUDA development on my alienware m11xR2 laptop (the one with the dreaded optimus…i.e. no linux support from nvidia…). Actually, I bought it for work and I NEED to do some CUDA on it…

It has the NVIDIA Geforce 335M which I am pretty sure has CUDA support.

I have Win7-64 bit OS.

I have downloaded and tried to install the latest developer driver, devdriver_3.2_winvista-win7_64_260.61_notebook, but when I try to install it, it says that no supported hardware can be found.

What are my options? Am I trying to install the wrong driver?

Surely this is possible. I mean… I know linux isn’t supported but this is Windows 7…

CD

Yes, install the release and not the developer drivers ( the ones accessible directly from the NVIDIA home page ). You should have no problems running CUDA programs on this laptop, I have one myself. It is worth reading the developer notes for Optimus platforms, however.

regards,

Andrew.

Yes, install the release and not the developer drivers ( the ones accessible directly from the NVIDIA home page ). You should have no problems running CUDA programs on this laptop, I have one myself. It is worth reading the developer notes for Optimus platforms, however.

regards,

Andrew.

Incidentally, CUDA will work on the Alienware11xR2 with Linux, but I have experienced permissions issues. ie, the SDK samples must be initially run as root in oder to work. I’m waiting for a reply from NVIDIA on the explanation/solution for this.

Regards,

Andrew.

Incidentally, CUDA will work on the Alienware11xR2 with Linux, but I have experienced permissions issues. ie, the SDK samples must be initially run as root in oder to work. I’m waiting for a reply from NVIDIA on the explanation/solution for this.

Regards,

Andrew.

I have an alienware M11x R2 and I am having a bad time trying do install cuda and made it work, i.e. compile and have the executable files to recognize the NVIDIA device. How did you manage to have cuda working?

Thanks for any light.

Tarcisio

I have an alienware M11x R2 and I am having a bad time trying do install cuda and made it work, i.e. compile and have the executable files to recognize the NVIDIA device. How did you manage to have cuda working?

Thanks for any light.

Tarcisio

[list=1]

[*] Install the 260.19 developer drivers. Opt not to modify your X config ( this will not work as the driver does not support using the NVIDIA card as a display device )

[*] Install the nvidia module with modprobe ( may or may not be necessary ).

[*] Install the CUDA Toolkit. Modify your startup file (.bashrc or whatever) to add the location of the cuda libraries to LD_LIBRARY_PATH

[*] Install the SDK

[*] Compile the cuda utils and deviceQuery sample in the SDK. If this doesn’t work you’ve done something wrong.

[*] run deviceQuery as root.

[list=1]

[*] Install the 260.19 developer drivers. Opt not to modify your X config ( this will not work as the driver does not support using the NVIDIA card as a display device )

[*] Install the nvidia module with modprobe ( may or may not be necessary ).

[*] Install the CUDA Toolkit. Modify your startup file (.bashrc or whatever) to add the location of the cuda libraries to LD_LIBRARY_PATH

[*] Install the SDK

[*] Compile the cuda utils and deviceQuery sample in the SDK. If this doesn’t work you’ve done something wrong.

[*] run deviceQuery as root.

Thanks, now I can run CUDA programs. By the way, it sufices to run deviceQuery once as root and then CUDA works ok for normal user. Quite curious.

Tarcisio

Thanks, now I can run CUDA programs. By the way, it sufices to run deviceQuery once as root and then CUDA works ok for normal user. Quite curious.

Tarcisio

Did anybody got any feedback from nvidia regarding this issue?

Also when the laptop goes on standby, that completely shut down cuda, nothing works anymore. I have to rmmod nivia, modprobe nvidia and rerun the devuiceQuery as a root.

Anybody has a better work around ?

I’m hoping planetmarshall could elaborate on how he got the cuda drivers installed? I’m on ubuntu and after installing drivers, I

try to startx and I get an xorg error that no screens are found? How did you get it working?

Thanks so much!

NEVER MIND… just read your post and realized you said NOT to modify the xorg

I completed the installation, and it passed the deviceQuery test.

But it cant execute the volumeRender or the fluidsGL executable,probably because I haven’t replaced the Xorg file, but if I do then the computers bugs itself. So to get it straight CUDA is allowed as long as there is no GL output?

Btw : The error I receive when I try to run the GL executables is this:

Xlib: extension “GLX” missing on display “:0.0”.
freeglut (./volumeRender): OpenGL GLX extension not supported by display ‘:0.0’

Thanks!!

I am running fedora 15 on an asus u36.

I did all of the above and I have a working CUDA installation (now except i didn’t install the 260.19 drivers but the 285.05 ones).

When I did that on Ubuntu the intel driver worked like a charm.

On fedora i am stuck with the vesa driver.

What can I do to enable the intel card?

I just want the nvidia card as a CUDA co-processor and the intel card as a display card.

Device query works but i am stuck in 1024*768 which is not very nice.

I have tried a costum xorg.conf but it doesn’t work.

Thank you in advance.

Apostolis