[Pw_forum] Do you think Intel I7 series could improve the parallel efficiency
vega lew
quantumdft at gmail.com
Wed Mar 25 18:38:18 CET 2009
Dear sir,
Thank you for your reply. I think discussing relationship of QE and hardware
is quite helpful.
Unfortunately, we don't have specific staff dealing with this matter. My
supervisor rarely show me 110,000 yuan (equal to ~$20,000). and tell me he
want the calculation environment promoted greatly.
So I must play the part of that staff. One year ago, I just finished a small
system (8 computers with Q6600 and 8G memory on intel S3000AH board / node )
and conquered the problems as you mentioned, such as motherboard, bios, dual
channel memory, OS, etc. I'm quite confident that I could make it this time.
This time, our aim is to establish a small DIY server standard platform
which could calculate less than 200 atoms effectively using QE/CPMD.
Unfortunately, we still have no infiniband or myrinet. So I think a strategy
of duo-socket/node (8 cores per node) is more suitable for us. as you just
told me, the bandwidth of the memory is key issue for us (we didn't concern
about it previously). and could you give some benchmarks of QE/CPMD running
on various platforms ( both server standard and PC standard). And is there
some suggestion for us to built a small cluster suitable for QE/CPMD with
less than $20,000(US dollar).
thank you for reading.
vega
On Wed, Mar 25, 2009 at 11:22 PM, Axel Kohlmeyer <
akohlmey at cmm.chem.upenn.edu> wrote:
> On Wed, 2009-03-25 at 22:51 +0800, vega lew wrote:
> > Dear sir,
> >
> > You answer is quite helpful to me. Thank you very much. Actully, the
> > speed of calculation for a large system is more
> > important for us. So I think the bandwidth is the key issue.
> >
> > could I recognize the bandwidth you mentioned as the FSB (Front system
> > bus) of the montherboard and the frequency of memory such as 800MHz or
> > 1066 MHz and so on?
>
> not really. you can see the bandwidth reliably from running the
> stream benchmark. those numbers do - of course - exist, but are
> currently under NDA, so nobody can talk about it until the end
> of next week, IIRC.
>
> from what is public already i can tell you, that the architecture
> of nehalem EP also requires that you pay great care as to where
> memory modules are plugged in. good mainboards will have memory
> slots in multiples of three. by placing memory modules in the wrong
> slots, you can reduce the bandwidth by a factor of three. ;)
>
> as with any new architecture, i would also expect that you'll have
> expect a significant amount of problems with mainboard, bios, drivers
> etc. and thus you should make sure that you have staff able to deal
> with it.
>
> axel
>
> > thank you for your reply
> >
> > vega
> >
> > On Wed, Mar 25, 2009 at 8:45 PM, Axel Kohlmeyer
> > <akohlmey at cmm.chem.upenn.edu> wrote:
> > On Wed, 2009-03-25 at 19:30 +0800, vega lew wrote:
> > > Dear all,
> > >
> > > Is there someone perform QE on intel I7 platform? Do you
> > think
> > > parallelzation QE could benefit from the QPI technology of
> > intel I7
> > > series very much?
> >
> >
> > the integrated memory controller and parallelization have very
> > little connection directly. please don't "believe" in
> > buzzwords.
> >
> > what _will_ help codes like Q-E in general is the greatly
> > enhanced
> > memory bandwidth, particularly the for larger systems. for
> > smaller
> > systems that mostly fit into the cache, the gains will be much
> > smaller.
> > the floating point unit of those cpus is essentially the same
> > as in
> > the current xeon processors.
> > since the various tests posted on the list here demonstrated
> > that
> > (some of) Q-E's parallelization schemes are very much affected
> > by
> > insufficient memory bandwidth, there will be a gain. however,
> > those
> > CPUs will also have a different cache architecture, which may
> > in
> > turn chew up some of the benefits, particularly for smaller
> > problems.
> >
> > cheers,
> > axel.
> >
> >
> >
> > >
> > > vega
> > >
> > > --
> > >
> >
> ==================================================================================
> > > Vega Lew ( weijia liu)
> > > PH.D Candidate in Chemical Engineering
> > > State Key Laboratory of Materials-oriented Chemical
> > Engineering
> > > College of Chemistry and Chemical Engineering
> > > Nanjing University of Technology, 210009, Nanjing, Jiangsu,
> > China
> > >
> >
> ******************************************************************************************************************
> > > Email: vegalew at gmail.com
> > > Office: Room A705, Technical Innovation Building, Xinmofan
> > Road 5#,
> > > Nanjing, Jiangsu, China
> > >
> >
> ******************************************************************************************************************
> >
> > > _______________________________________________
> > > Pw_forum mailing list
> > > Pw_forum at pwscf.org
> > > http://www.democritos.it/mailman/listinfo/pw_forum
> >
> > --
> >
> =======================================================================
> > Axel Kohlmeyer akohlmey at cmm.chem.upenn.edu
> > http://www.cmm.upenn.edu
> > Center for Molecular Modeling -- University of
> > Pennsylvania
> > Department of Chemistry, 231 S.34th Street, Philadelphia, PA
> > 19104-6323
> > tel: 1-215-898-1582, fax: 1-215-573-6233, office-tel:
> > 1-215-898-5425
> >
> =======================================================================
> > If you make something idiot-proof, the universe creates a
> > better idiot.
> >
> > _______________________________________________
> > Pw_forum mailing list
> > Pw_forum at pwscf.org
> > http://www.democritos.it/mailman/listinfo/pw_forum
> >
> >
> >
> > --
> >
> ==================================================================================
> > Vega Lew ( weijia liu)
> > PH.D Candidate in Chemical Engineering
> > State Key Laboratory of Materials-oriented Chemical Engineering
> > College of Chemistry and Chemical Engineering
> > Nanjing University of Technology, 210009, Nanjing, Jiangsu, China
> >
> ******************************************************************************************************************
> > Email: vegalew at gmail.com
> > Office: Room A705, Technical Innovation Building, Xinmofan Road 5#,
> > Nanjing, Jiangsu, China
> >
> ******************************************************************************************************************
> > _______________________________________________
> > Pw_forum mailing list
> > Pw_forum at pwscf.org
> > http://www.democritos.it/mailman/listinfo/pw_forum
>
> --
> =======================================================================
> Axel Kohlmeyer akohlmey at cmm.chem.upenn.edu http://www.cmm.upenn.edu
> Center for Molecular Modeling -- University of Pennsylvania
> Department of Chemistry, 231 S.34th Street, Philadelphia, PA 19104-6323
> tel: 1-215-898-1582, fax: 1-215-573-6233, office-tel: 1-215-898-5425
> =======================================================================
> If you make something idiot-proof, the universe creates a better idiot.
>
> _______________________________________________
> Pw_forum mailing list
> Pw_forum at pwscf.org
> http://www.democritos.it/mailman/listinfo/pw_forum
>
--
==================================================================================
Vega Lew ( weijia liu)
PH.D Candidate in Chemical Engineering
State Key Laboratory of Materials-oriented Chemical Engineering
College of Chemistry and Chemical Engineering
Nanjing University of Technology, 210009, Nanjing, Jiangsu, China
******************************************************************************************************************
Email: vegalew at gmail.com
Office: Room A705, Technical Innovation Building, Xinmofan Road 5#, Nanjing,
Jiangsu, China
******************************************************************************************************************
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.quantum-espresso.org/pipermail/users/attachments/20090326/4a45205e/attachment.html>
More information about the users
mailing list