From: Sunny (sunny.day.programmer.2019_at_gmail.com)
Date: Tue Jun 25 2019 - 02:06:25 CDT
Dear Bryan
Thanks for the kind reply, Unfortunately I don't have access to a setup
which I own right now but I will keep your advice in my mind regarding
using Linux. But the question is why the windows version that the NAMD
developers put on the website is faulty (if I am even right about being
faulty, I think I need to setup the OS in some special way maybe)? I
appreciate their hard work to deliver this great software to community but
I think they should take responsibility and at least answer a solution to
the problem.
Huge thanks for the response
Cheers
Sunny
On Tue, Jun 25, 2019 at 2:36 AM Bryan Roessler <bryanroessler_at_gmail.com>
wrote:
> Hi Sunny,
>
> I briefly ran your simulation setup on my dual-Xeon rig with a couple of
> GTX GPUs running Fedora 30 and attached the output.
>
> So I am assuming that this is a Windows-specific issue?
>
> If so, I would recommend to all of you that you check your Windows
> Defender settings to exclude the namd2 binary and your simulation
> directories. And to be quite succinct, perform your simulations on a Linux
> OS so you don't have Windows "features" mucking with your production work.
>
> Cheers,
> Bryan
>
>
>
> On Sat, Jun 22, 2019 at 10:44 PM Sunny <
> sunny.day.programmer.2019_at_gmail.com> wrote:
>
>> Dear Maximilian
>> I have the same issue as mentioned in previous listings, "*FATAL ERROR:
>> High global CUDA exclusion count!*
>> <https://www.ks.uiuc.edu/Research/namd/mailing_list/namd-l.2018-2019/2028.html>"
>> but no one answered. the slow performance is due to a unknown reason which
>> the system collapse under its own force to a very small and unnatural
>> arrangement abruptly after the energy minimization step. this will increase
>> the neighbor list and force computations and hence the slow performance.
>> you can check this by getting dcd output from the simulation every single
>> frame and watch the evolution of the system. for me it happens in a couple
>> of hundred to thousands steps. I reached out the listing to ask for any one
>> with experience on the matter with no response yet. I hope someone can help
>> both of us on this. all of the files required for my simulation are put in
>> the link below:
>> https://www.dropbox.com/s/feigmwjwuqu99ts/testCUDA.rar?dl=0
>>
>> bests regards
>> Sunny
>>
>> On Thu, Jun 20, 2019 at 12:14 PM Maximilian Ebert <mebert_at_chemcomp.com>
>> wrote:
>>
>>> HI there,
>>>
>>> The CUDA acceleration on our Windows 10 desktop computer seems not to
>>> work. The NVIDIA P5000 GPU idles around 1-5% for small or large systems. We
>>> have a Linux equivalent with a P4000 and all appears to work fine. I
>>> attached the GPU stdout from windows for 1 CPU + 1 GPU and 4 CPUs + 1 GPU.
>>> Both run slow. Any idea what is going on?
>>>
>>> https://drive.google.com/file/d/1-CubaGM0URZeWUto6HbodsS7UE-Lq-fI
>>> https://drive.google.com/file/d/1-Fnsp7i7saNfzZwoQQ-eXB5s31q-Atw7
>>>
>>> Thanks,
>>> Max
>>>
>>>
>>> --
>>> Chemical Computing Group | www.chemcomp.com <http://www.chemcomp.com> |
>>> 910-1010 Sherbrooke W., Montreal QC H3A 2R7, Canada | +1 514 393 1055
>>>
>>> Manage Communications: www.chemcomp.com/Manage_Subscription.htm
>>> <http://www.chemcomp.com/Manage_Subscription.htm>
>>>
>>>
This archive was generated by hypermail 2.1.6 : Thu Dec 31 2020 - 23:17:11 CST