Re: [AMBER] Error in Installing Amber 20 on GPU

From: David A Case <>
Date: Fri, 3 Jul 2020 09:37:24 -0400

On Fri, Jul 03, 2020, Suchetana Gupta wrote:

>Thanks for the response. As per your suggestion, we edited the run cmake
>file at DCUDA_TOOLKIT_DIR ta cuda-10.0. The installation was okay. But the
>make test.cuda.serial gave error. Attaching it in the txt file. Also,
>attached is the output for nvcc --output.

>Now in our machine, on typing nvidia-smi, we are getting the error: failed
>to initialize nvml: driver library version mismatch

I don't see how just changing an option in run_cmake could have caused
nvidia-smi to change its behavior. Most of your test errors come from
things that seem related to this: not finding a CUDA device, or not
being able to intialize it.

This is not something that can be debugged remotely: you will need to
find someone who has access to your machine and can examine the problem
of the mismatch between drivers and library version.


AMBER mailing list
Received on Fri Jul 03 2020 - 07:00:03 PDT
Custom Search