Installation and Various

Good evening to everyone, I’m a new user of CUDA technology, or at least I’d like become one.
I have some -quite- stupid questions to ask.
I apologies if this is not the right forum section, I hope the moderators will move it in the proper one.
Let’s start:-

  1. I’m a Ph.D student of mathematical finance, and I need the power of my GPU for perform some complex numerical simulation -Montecarlo primarily -. So I presume I’ll need the Nvidia CUDA Toolkit 8, right?
    Or should I install some other things?

2)The CUDA Toolkit 8 -or the other things that I should install- provides an environment such as an IDE or I should install the IDE for CUDA separately? -If yes can you suggest one?-

3)I use RHEL 7.2 workstation, now in the Nvidia install section, disabling Nouveau driver seems not a mandatory action since at the point 3.2.8) I’m redirect to perform the post-installation action, but in the red hat forum this procedure seems mandatory. So, is there a way to perform the installation without touch the Nouveau driver? This is not a effective limit, but I prefer touch as less as possible in the OS.

4)In the installation guide seems that clang 3.8 is required, but I’m not able to find such version for my OS, is clang a mandatory package or it is an alternative respect gcc ? -since after, the guide, asks only gcc-

Machine details:
PC : Dell workstation t7910
GPU : Nvidia Quadro M5000
OS : Red Hat Enterprise Workstation 7.2

Thanks to everyone.
All the Best.

Some quick responses… others can provide more detail later if need be:

  1. CUDA Toolkit 8 is supported under RHEL, so yes, that’s fine. It comes with examples that you can compile with gcc

  2. Nsight Eclipse Edition is provided: https://developer.nvidia.com/nsight-eclipse-edition

  3. Disabling Nouveau is required, as to provide CUDA support you will need to install the native NVIDIA display driver that is compatible with the CUDA Toolkit version you install.

  4. I have compiled source files with gcc in Linux just fine, clang is supported, but not necessary.

Dear vacaloca, thank you for the reply. I have just two more questions.

1)Assuming that Eclipse Nsight is “contained” in the Toolkit -and so I’ll not need to install it separately-, I’d like to know if for install the CUDA Toolkit, some dependencies are required -such as,just for make two examples , CUDA library, CUDA headers… - or if the Toolkit is self-contained.

2)What do you mean with:-

as to provide CUDA support you will need to install the native NVIDIA display driver that is compatible with the CUDA Toolkit version you install.

which point is it in the guide?

Sorry for my quite basic question.
Thank you anyway for your previous reply.
All the best.

GiPa

  1. The CUDA Toolkit includes NVIDIA drivers that are compatible with the version that you download. The most recent version of any NVIDIA driver will be compatible with the current version of the CUDA toolkit, should you install the video driver first. Depending on the method of installation yes, some dependencies and libraries may be required. For that you can Google further.

  2. To access CUDA features, the NVIDIA driver needs to be installed on your system… as simple as that. Not sure where the installation guide mentions it honestly, as I haven’t looked in a while… but that is the case.

Dear vacaloca, thank you for your reply again. Now is everything clear. Obviously, the NVIDIA basic driver for the correct behaviour of the GPU must be present in the pc before installing CUDA.
I believed you were referring to some other CUDA external driver.
The last question, and we have done.
Where I can find the NVIDIA-driver-version required for the CUDA-toolkit?

Thank you again for the time dedicated to me and the kind attention.
All the best.

The CUDA toolkit includes a compatible driver by default as part of the installation, just follow the installation instructions for that.