My IT dept controls the Tesla I work on. I asked them to install Cuda 2.0, which they did, but they apparently did not install the driver. So that was installed yesterday, but I still cannot use textures which I was expecting to with the new driver.
If the new driver is installed does the SDK, profiler etc have to be re-installed?
Un-install all CUDA stuff…and then install “driver”, “Toolkit” followed by the SDK for CUDA 2.0 in the same order listed above.
There’s no dependency chain in the order in which you install the three.
@netllama:
but the ‘cuda’ homepage tells the users to install the three installations in order?
ref: [url=“CUDA Toolkit 11.7 Update 1 Downloads | NVIDIA Developer”]http://www.nvidia.com/object/cuda_get.html[/url]http://www.nvidia.com/object/cuda_get.html
The implication is that you can’t use them unless they are installed in the order listed. For HOSTEMU, you don’t’ even need the driver, so the toolkit certainly doesn’t have a requirement to install the driver first.
Is it possible to install the SDK and toolkit on a machine that doesn’t have a CUDA-enabled device? I though I read somewhere that one of the installers checks for CUDA and won’t install if you don’t have a CUDA device in the system.
I’d like to be able to do some CUDA coding on my laptop (which doesn’t have a CUDA-compatible GPU), even if it means that I have to run everything in emulation mode.
Also, as a testing thing…it would be really cool if nVidia made a ‘fake’ device driver for emulation in the same way Daemon Tools gives your computer a virtual CD drive. If we could install this device driver and ‘fool’ the system into thinking it had a CUDA device, and we could select what type of device it was, it would be a great way to test code for deployment on non-development machines, as well as for checking how the code runs on different devices (think 8800GT vs. C1060, etc.).
Yes, you can.