Re: NVIDIA Quadro(R) GV100 card for NAMD studies

From: Giacomo Fiorin (giacomo.fiorin_at_gmail.com)
Date: Mon Nov 05 2018 - 07:29:48 CST

Hello Sesha, you did not ask a question specific enough to have an answer
in the first place. The number of atoms, number of steps, cutoffs, I/O
frequencies and last but not least the integration time step are all
variables that affect dramatically computational speed.

Also, since you are asking people to volunteer information, you could start
by sharing the performance that you measured for your simulation and
hardware. This may encourage people to participate in a discussion more
than just asking to optimize your particular simulation.

Giacomo

On Mon, Nov 5, 2018 at 1:58 AM sesha surya vara prasad reddy karri <
prasadreddy2792_at_gmail.com> wrote:

> Hello Giocomo,
>
> No one has replied for this mail.. Can i get the answer for this?
> Typically how much time is taken to complete a NAMD job in GV100 nvidia
> card?
>
>
>
>
> ‌
> [image: Mailtrack]
> <https://mailtrack.io?utm_source=gmail&utm_medium=signature&utm_campaign=signaturevirality4&> Sender
> notified by
> Mailtrack
> <https://mailtrack.io?utm_source=gmail&utm_medium=signature&utm_campaign=signaturevirality4&>
>
> On Fri, Nov 2, 2018 at 5:35 PM sesha surya vara prasad reddy karri <
> prasadreddy2792_at_gmail.com> wrote:
>
>> Thanks for valuable information Giocomo,
>>
>> This GPU is really helping me because my system is large. i know that
>> time required to complete a job is dependent on our md system. but i want
>> to know how much time it takes to complete NAMD job of 10ns of any other
>> systems GV100 nvidia card. Thank you
>>
>> Regards,
>>
>> ‌
>> [image: Mailtrack]
>> <https://mailtrack.io?utm_source=gmail&utm_medium=signature&utm_campaign=signaturevirality4&> Sender
>> notified by
>> Mailtrack
>> <https://mailtrack.io?utm_source=gmail&utm_medium=signature&utm_campaign=signaturevirality4&>
>>
>> On Fri, Nov 2, 2018 at 5:19 PM Giacomo Fiorin <giacomo.fiorin_at_gmail.com>
>> wrote:
>>
>>> Hi Sesha, it is generally not possible to use all of the cores of a GPU
>>> simultaneously, since they implement very different types of calculations.
>>> This is especially true for specialized circuits such as the tensor cores
>>> of Volta-generation GPUs and the ray-tracing cores of the Turing ones. You
>>> should look only at the cores that the MD simulation programs use: single-
>>> and double-precision floating-point cores.
>>>
>>> It is also nearly impossible to keep all cores of one type working all
>>> the time, because loading data on/off takes a significant chunk of time--0000000000003fc2680579eae3f0--

This archive was generated by hypermail 2.1.6 : Sun Sep 15 2019 - 23:20:05 CDT