Hi Murat,
Yes that number is WITH hydrogen mass repartitioning - it's actually
constant pressure as well, their run is constant volume but I didn't have
the NVT number to hand right now but that would for sure be faster than
the NPT numbers I gave you.
If you look on
http://ambermd.org/gpus/ you will also see benchmark
numbers for more conventional runs with 2fs as well.
You might also want to watch the webinar we had yesterday about AMBER 14:
http://www.gputechconf.com/resources/gtc-express-webinar-program - It was
recorded so I think a link to it should be on this page in a few days. It
includes Adrian talking about his testing of HMR if you are interested.
With regards to comments in namelists - sorry didn't see that before
(don't mix topics in emails :-P ).
You use a ! for the comment. E.g.:
&cntrl
ntx = 1, ! Read in formatted coords, no vels.
ntb = 0 ! No periodic box.
cut = 1000.0, ! Nonbonded cutoff.
...
...
/
We should probably document that somewhere or update the tutorials to
include such comments.
All the best
Ross
On 5/14/14, 12:41 PM, "MURAT OZTURK" <murozturk.ku.edu.tr> wrote:
>Oh, learning new things everyday. I actually subscribed to the maillist
>when that particular thread was halfway thrugh. I have the last few post
>in
>my inbox, did not make much sense when I received them. :)
>
>Going back to the original subject, reported 239.14ns/day with amber is
>also WITH hydrogen mass repartitioning. Match and win. Cool.
>
>I don't think I will actually try that, though, it is very nice to know.
>
>Now, about this comments-in-namelist thing..? No commenting, really?
>
>Regards,
>
>Murat
>
>
>On Wed, May 14, 2014 at 10:33 PM, Jason Swails
><jason.swails.gmail.com>wrote:
>
>> On Wed, 2014-05-14 at 22:22 +0300, MURAT OZTURK wrote:
>> > No, you need not say more (performance wise). :)
>> >
>> > Actually I was curious about the inner workings. The page you linked
>>also
>> > mentions 4 fs timestep. This is not recommended in amber (nor
>>gromacs) as
>> > far as I know. I mostly do 2 fs, sometimes even 1 fs timestep. If
>>their
>> > ns/day measurement is based on that, then it is off by %50. If they do
>> > actually pull off a 4 fs timestep, then how?
>>
>> Hydrogen mass repartitioning -- they make the hydrogen atoms heavier and
>> their attached heavy atoms lighter. For instance, a methyl group will
>> go from
>>
>> 1.01
>> |
>> 1.01 -12.01 - 1.01
>>
>> to
>>
>> 3.00
>> |
>> 3.00 -6.01 - 3.00
>>
>>
>> so that the total mass of the functional group stays the same but the
>> heavier hydrogens vibrate with a (much) lower frequency. You still need
>> to use SHAKE, but with the heavier mass the deviations are small enough
>> for SHAKE to correct in a sensible number of iterations.
>>
>> Water constraints are solved analytically (SETTLE), so repartitioning is
>> unnecessary there.
>>
>> You can do repartitioning in Amber using the ParmEd program (see the
>> manual for details on the HMassRepartition command).
>>
>> Hope this helps,
>> Jason
>>
>> --
>> Jason M. Swails
>> BioMaPS,
>> Rutgers University
>> Postdoctoral Researcher
>>
>>
>> _______________________________________________
>> AMBER mailing list
>> AMBER.ambermd.org
>> http://lists.ambermd.org/mailman/listinfo/amber
>>
>_______________________________________________
>AMBER mailing list
>AMBER.ambermd.org
>http://lists.ambermd.org/mailman/listinfo/amber
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Wed May 14 2014 - 13:00:06 PDT
