[Wrf-users] More cores or faster processors?

[Russel Morison]

Hi Nick,

             so this is a tricky question. It really depends on your model runs, and what causes you speed limiter.

You will notice that each of these cpus has the same cache and memory bandwidth, but different cache sizes,

and cache sizes per core (the E5-2667 has 3.125 MB/core, 8 cores and the E5-2683 has 2.5 M bytes/core 16 cores).

but the E5-2683 has 40 M bytes of cache vs 25 M cache.

So if, when you have divided you domain up into sub domains, 128 sub domains for the 2667s with 400M of cache total,

versus 256 sub domains for the 2683s with 640 M of cache, I expect the extra cache will be very beneficial.

Also the percentage of your domain per M byte of cache is smaller with the 2683s vs the 2667s (100/640 vs 100/400).

So I expect you will get less cache flushing with the 2683s, unless you domains are small and the domain overlap becomes

significant. Also the total compute power in each 2683 is about 30% more.


For most model runs I expect the 2683's would be faster.


Russel .

________________________________
From: wrf-users-bounces at ucar.edu <wrf-users-bounces at ucar.edu> on behalf of Witcraft, Nick <nick.witcraft at ncdenr.gov>
Sent: Tuesday, 10 May 2016 5:48:10 AM
To: wrf-users at ucar.edu
Subject: [Wrf-users] More cores or faster processors?

Hi,

We are in the process of building a new cluster, with a total of 8 compute nodes. Would it be better to go with a faster processor with less cores, or a slower processor with more cores?
Sample options, very similar price:
Intel(r) Xeon(r) E5-2667 v4 3.2GHz,25M Cache,9.60GT/s QPI,Turbo,HT,8C/16T (135W) Max Mem 2400MHz
Intel(r) Xeon(r) E5-2683 v4 2.1GHz,40M Cache,9.60GT/s QPI,Turbo,HT,16C/32T (120W) Max Mem 2400MHz

I did look at Q13 of the FAQ ( http://www2.mmm.ucar.edu/wrf/users/FAQ_ ... ation.html<http://www2.mmm.ucar.edu/wrf/users/FAQ_files/FAQ_wrf_installation.html> ), but it doesn't really answer this particular question.


Thank you,
Nick Witcraft
North Carolina DAQ

Nicholas Witcraft, Meteorologist II
NC DEQ, Division of Air Quality
Planning Section, Attainment Planning Branch
1641 Mail Service Center
Raleigh, NC 27699-1641
Phone/Fax: 919-707-8484
Email:  Nick.Witcraft at ncdenr.gov<mailto:Nick.Witcraft at ncdenr.gov>
http://www.ncair.org<http://www.ncair.org/>
[cid:image001.png at 01D1058E.93ED1C60]
Email correspondence to and from this address is subject to the
North Carolina Public Records Law and may be disclosed to third parties.

-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mailman.ucar.edu/pipermail/wrf-users/attachments/20160510/6f5b4b7b/attachment-0001.html 
-------------- next part --------------
A non-text attachment was scrubbed...
Name: image001.png
Type: image/png
Size: 11798 bytes
Desc: image001.png
Url : http://mailman.ucar.edu/pipermail/wrf-users/attachments/20160510/6f5b4b7b/attachment-0001.png