Random crashes after a true CUDA crash

Has anyone else been experiencing random crashes?
I’m using WinXP, and CUDA 1.0.
The WinFast 8800 GTX card I’m using works fine before I install the CUDA 1.0 driver. But recently it starts crashing randomly, once on a pure DX9 program, once on a simple OpenGL PBO test program (that uses a trivial shader), and numerous times on my CUDA program.
The random crashes happen when GPU related programs are launched after a true crash of a CUDA program. The randomly crash behavior doesn’t go away after reboot (I once shut down power for an entire day, GPU temperature gets down to 58, and that DX9 program still crashes), but it does go away if the display turns dark for about a second and back (suspected to be a driver reset).

I have experienced these random crashes, when trying to use VBOs to output graphics based on data generated with CUDA. When I disable the CUDA component, the program runs fine.

When I just run with the GL component, everything is fine. When I link the too, crashes

occur unpredictably. I am running with CUDA 1.0, on linux with the latest drivers.

Gordon

My program also outputs graphics using CUDA generated VBO.
I experienced loads of GL crashes in 0.8. But when I decreased number of VBOs used to 1, everything went fine then.
Maybe GL interpolability is still instable?

Hard to know. I am using only a single VBO.

Don’t those nVidia guys ever come here?
It’s becoming worse.
Now my program ALWAYS crash, unless I enable some debug code which totally cancels out GPU’s advantage.
CUDA would be useless for us if this is not fixed soon.

Found a GLSL vertex shader that guarantees a crash now on my card.

void main()

{

	gl_Position = ftransform();

	//gl_FrontColor = vec4(gl_Vertex.xyz * 5.0+0.5);//doesn't crash

	gl_FrontColor = vec4(gl_Vertex.zzz * 5.0+0.5, 1.0);//crash

}