From: Kanishk Jain (kjain_at_ucla.edu)
Date: Wed Mar 08 2017 - 19:12:16 CST

Axel, thanks for the advice!

Joao, your question prompted me to update my CUDA driver to 8.0.63 and my
NVIDIA driver is 367.15.10.35f01. After updating the CUDA driver, I tried
my simulation again and I didn't get that error anymore. I'm assuming this
means that my GPU was recognized and is being used now?

-Kanishk

On Wed, Mar 8, 2017 at 1:32 PM, João Ribeiro KS <jribeiro_at_ks.uiuc.edu>
wrote:

> Dear Kanishk,
>
> can you tell us the CUDA Driver Version that you installed and the GPU
> Driver Version? You can find this in your System Preferences/CUDA.
>
>
> Best
>
> Joao
>
> On Wed, Mar 8, 2017 at 3:26 PM Axel Kohlmeyer <akohlmey_at_gmail.com> wrote:
>
>> On Wed, Mar 8, 2017 at 3:48 PM, Kanishk Jain <kjain_at_ucla.edu> wrote:
>>
>> I double checked and noticed there was indeed an initial error which may
>> explain why I'm not able to connect to NAMD:
>>
>> Reason: FATAL ERROR: CUDA error cudaGetDeviceCount(&deviceCount) in file
>> src/DeviceCUDA.C, function initialize
>>
>> on Pe 3 (Kanishks-iMac.local): CUDA driver version is insufficient for
>> CUDA runtime version
>>
>> Does this mean that my GPU (NVIDIA GeForce GT 750M 1024 MB) isn't good
>> enough? I updated the NVIDIA web drivers as well, but still nothing.
>>
>> I've used the non-CUDA version of NAMD (NAMD_2.12_MacOSX-x86_64-multicore)
>> succesfully, but I was hoping that the CUDA version would allow me to
>> utilize my GPU and make the MD calcs much faster.
>>
>>
>> ​i cannot comment on the driver issue, but i suggest you carefully study
>> the NAMD release notes which should reference GPU compatibility.
>>
>> ​also, ​you seem to have quite inflated expectations about how much
>> speedup you can get from your ​GPU. your GPU appears to be a laptop model
>> with a limited number of CUDA cores. it also will have a lower clock and
>> heat budget than high-end desktop models. so i would be surprised, if you
>> get more than 20% of the CUDA performance of a high-end desktop or server
>> GPU. if you are also going to share the GPU across multiple CPUs, there is
>> not much speedup left.
>>
>> ​axel.​
>>
>> -Kanishk
>>
>> On Tue, Mar 7, 2017 at 3:08 PM, Kanishk Jain <kjain_at_ucla.edu> wrote:
>>
>> Hello all,
>>
>> I'm an MD novice, starting out on QwikMD. I'm using a mac with the
>> following versions of VMD and NAMD, respectively: vmd193macx86nocuda
>> and NAMD_2.12_MacOSX-x86_64-multicore-CUDA.
>>
>> My problem is that when I try to start a simulation, I keep getting the
>> following error:
>>
>> ERROR) Error connecting to localhost on port 3000
>>
>> I've seen a similar question to mine as it applies to Linux machines, but
>> as I'm a novice, I'm not quite sure how to implement the solutions
>> mentioned in those threads to my situation. Any help would be appreciated!
>>
>> -Kanishk
>>
>> --
>> Sincerely,
>> Kanishk Jain
>> Steven G. Clarke Laboratory, PhD Candidate
>> Biochemistry and Molecular Biology Program
>> Department of Chemistry and Biochemistry
>> University of California, Los Angeles
>>
>>
>>
>>
>> --
>> Sincerely,
>> Kanishk Jain
>> Steven G. Clarke Laboratory, PhD Candidate
>> Biochemistry and Molecular Biology Program
>> Department of Chemistry and Biochemistry
>> University of California, Los Angeles
>>
>>
>>
>>
>> --
>> Dr. Axel Kohlmeyer akohlmey_at_gmail.com http://goo.gl/1wk0
>> College of Science & Technology, Temple University, Philadelphia PA, USA
>> International Centre for Theoretical Physics, Trieste. Italy.
>>
> --
> ……………………………………………………...
> João Vieira Ribeiro
> Theoretical and Computational Biophysics Group
> Beckman Institute, University of Illinois
> jribeiro_at_ks.uiuc.edu
> +1 (217) 3000380 <(217)%20300-0380>
>

-- 
Sincerely,
Kanishk Jain
Steven G. Clarke Laboratory, PhD Candidate
Biochemistry and Molecular Biology Program
Department of Chemistry and Biochemistry
University of California, Los Angeles