I’ve just built a new machine with tesla cards, a tyan s7025 motherboard etc…
Anyway I’ve installed everything under Ubuntu 9.04 and I’m having a few problems.
When I run deviceQuery in the sdk it tells me there is no device and I’m in emulation mode???
Also most of the sdk examples throw the following error (I use the Mandelbrot program as an example)
[ CUDA Mandelbrot & Julia Set ]
Initializing GLUT...
Xlib: extension "GLX" missing on display ":0.0".
freeglut (./Mandelbrot): OpenGL GLX extension not supported by display ':0.0'
I’m using the onboard graphics card for my display (which is an aspeed ast2050 chipset) I’m fairly sure this is the root of the problem as it’s clearly not set up correctly but I have no idea how to do that. As the tyan board is mentioned in the NVIDIA how to build a PSC then I know it must be possible.
Probably don’t have a CUDA capable driver installed.
That card is basically a 2D framebuffer intended only for console work, and the (closed source) driver doesn’t support OpenGL. So none of the OpenGL examples will work. If you want OpenGL, then you will have to get yourself a suitably capable card and install it.
You might have to disable the onboard video. It is quite possible that the BIOS won’t load the Tesla VGA BIOS if the onboard video is active. Older Tyan Opteron boards I used were like that.
Ok, that sounds possible. But how do I install the tesla driver if I have no video output to see what I’m doing? sorry if the answer to this is basic, but I’m relatively new to linux…
Over a network is probably the easiest way. Or a serial console. That Aspeed gpu on the board has some pretty sophisticated KVM and remote console over ethernet facilities as well.