[TRNSYS-users] Quicker processing speeds?

keilholz, werner werner.keilholz at cstb.fr
Fri Nov 17 06:21:32 PST 2006


Interesting topic, thanks for bringing it up and all the comments. 
 
Just to complete the Simulation Studio side: as far as plug-ins go, they are
all started in a different thread which then even starts a new process (the
thread sleeps until the process terminates). 
So anything happening in a plug-in should suck at a new processor, and each
new plug in will grab a new processor as long as there is one available  -
at least in theory, because we've never even touched such a machine here.
 
Werner
 
  _____  

De : trnsys-users-bounces at engr.wisc.edu
[mailto:trnsys-users-bounces at engr.wisc.edu] De la part de Blair, Nate
Envoyé : lundi 13 novembre 2006 17:33
À : trnsys-users at engr.wisc.edu
Objet : Re: [TRNSYS-users] Quicker processing speeds?
 
I would just like to quickly add that I think the really critical piece for
using dual-core processors is the optimizing software. Is it sequential or
can it do a variety of runs all at the same time? I know that TRNSYS has
been used with a variety of "multi-run" tools such as Condor that send out
many runs of a software to many different computers. However, this only
works if the process is able to be parallelized (like in a monte-carlo
analysis). If the tool requires that the results of each run be analyzed
before the next run is initiated, then it must be sequential unfortunately. 
 
I've also found that a dual-core is helpful with many TRNSYS runs because
you can still use the "other core" for other tasks such as email, Excel,
etc. while TRNSYS works along in the background. 
 
I'm still waiting for a 4 Ghz machine myself. 
 
 
 
  _____  

From: trnsys-users-bounces at engr.wisc.edu
[mailto:trnsys-users-bounces at engr.wisc.edu] On Behalf Of Janne Paavilainen
Sent: Monday, November 13, 2006 7:35 AM
To: Charron, Rémi; trnsys-users at engr.wisc.edu
Subject: Re: [TRNSYS-users] Quicker processing speeds?
 
Hi!
 
OK, as you can see from the replies it's quite a messy jungle with PCs
nowadays. I've also been a few times in the position of trying to justify
the needs/benefits of a new work-station for simulations (not only TRNSYS)
and it's not always easy. Your interest in reducing sim time is
understandable as with your current setting you end up easily in >24hr
simulation times (thousands of 100sec sims).  
 
Trying to summarize some guide-lines from previous mails:
 
* If you have an older generation CPU, then upgrading to a good PC of the
current CPU generation will probably already halve your simulation time,
give or take. This you can see e.g. from Rafs test, or comparing bench-mark
results on net reviews.
 
* If you already have a PC with a current generation CPU, then you have to
start thinking about price vs. performance. Getting a 20-30% CPU performance
boost (sim time reduction) compared to good standard PCs might almost double
the price of your new work-station, this might be difficult to justify.
 
* Also, looking at the previous point: If you can divide your problem into
two, buying e.g. two good standard PCs instead one of the fastest possible,
might almost halve your total sim time instead of the 20-30% reduction for
one simulation for the same price. And after the project you have two
perfectly good PCs to give to new employees, instead of one.
 
* The benefit of a dual-core CPU is harder to comment on. You might be able
to do with one PC the same as with two computers as described above, but it
would be better to test it before buying. This is possible at least with
TRNSYS-only, but it might be more complicated in combination with other
programs. You should be able to test this with a single core CPU, on your
current PC (the simulation will be a crawl, but at least you should be able
to see if it works).
 
* Running two simultaneous simulations on dual-core generates a lot of heat,
take care that the cooling of the PC is well done. If you are running
simulations of several hours like this, you might end up crashing your hard
disk in no-time because of the elevated operating temperature. This problem
is not over-exaggerated!!!
 
* With a dual core CPU you can work normally while a TRNSYS simulation is
running in the background at full speed. This wasn't possible with single
core CPUs.
 
* RAM isn't an issue with TRNSYS, anything above 1GB is overkill, unless you
are using other software which need more (actually, already 1GB is overkill,
if only for TRNSYS+OS). 
 
* Hard disk isn't an issue either, modern HD:s can handle about any amount
of data that TRNSYS itself has to read and can spit out. (Unless you are
doing something else that's HD intensive at the same time)
 
 
 
A suggestion to TRNSYS development group:
It might be a good idea to have some general guide-lines about this on the
TRNSYS home pages in addition to the general hardware
requirements/recommendations. Generally, when a new project involving
simulations is started, almost always the need for a new computer is
discussed in the first meeting. The "Hardware requirements" section in the
manual does not give much help considering e.g. Rémis problem. This might
easily lead to "buying overkill" or spending money on the wrong part of the
hardware. And trying to google the www you end up spendings various hours
reading and still not knowing which way you should go.
 
 
 
Regards
Janne
 
----------------------------------------------------------------------------
-
Janne Paavilainen 
MSc, PhD student
Researcher in Energy and Environmental engineering
----------------------------------------------------------------------------
-
Solar Energy Research Center SERC
Dalarna University College
SE-781 88 Borlänge
Sweden
----------------------------------------------------------------------------
-
tel +46 (0)23 778728
fax +46 (0)23 778701
e-mail: jip at du.se <mailto:jip at du.se> 
----------------------------------------------------------------------------
-
www.du.se <http://www.du.se> 
www.serc.se <http://www.serc.se> 
www.eses.org <http://www.eses.org> 
www.uni-kassel.de/fb15/ite/solar/solnet/
<http://www.uni-kassel.de/fb15/ite/solar/solnet/> 
----------------------------------------------------------------------------
-
 
 
----- Original Message ----- 
From: Charron, Rémi <mailto:remi.charron at nrcan-rncan.gc.ca>  
To: trnsys-users at engr.wisc.edu <mailto:trnsys-users at engr.wisc.edu>  
Sent: Friday, November 10, 2006 6:43 PM
Subject: [TRNSYS-users] Quicker processing speeds?
 
 
Hello, 
I am currently running TRNSYS with using a computer with a 3.2 GHz prcoessor
with 512 MB of RAM.  It takes approximately 100 seconds to run a simulation.
This is a reasonable amount of time, except that I am running TRNSYS with an
optimisation program and am requiring to perform thousands of TRNSYS
simulations per run.  I am thinking of getting a quicker computer or
workstation to help speed TRNSYS up.  Has anyone had experience running
TRNSYS using duel core processors or other high performance computers or
workstations?  If so, what have been your experiences?  Can the computation
time be significantly reduced with the use of a more powerful computer?
Thanks, 
Rémi Charron 
Research Officer 
Photovoltaic and Hybrid System Program 
CANMET Energy Technology Centre 
1615 Lionel-Boulet Blvd, P.O. Box 4800, J3X 1S6 
Varennes, Quebec 
Tel: (450) 652-7948, Fax: (450) 652-5177 
 <http://cetc-varennes.nrcan.gc.ca> http://cetc-varennes.nrcan.gc.ca 

  _____  

_______________________________________________
TRNSYS-users mailing list
TRNSYS-users at engr.wisc.edu
https://www.cae.wisc.edu/mailman/listinfo/trnsys-users
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.onebuilding.org/pipermail/trnsys-users-onebuilding.org/attachments/20061117/5830f879/attachment-0001.htm>


More information about the TRNSYS-users mailing list