<div dir="ltr">   Thank you, Michael ! This will be useful too !<div style>   Nevertheless, we can not forget that, sometime ago, Prof. <span class="" style>Blaha</span> commented that the &quot;Cache Memory&quot; is more important than the &quot;clock&quot; for WIEN2K (discarding extreme cases, of course).</div>
<div style>   All the best,</div><div style>                   Luis</div></div><div class="gmail_extra"><br><br><div class="gmail_quote">2013/9/12 Michael Sluydts <span dir="ltr">&lt;<a href="mailto:michael.sluydts@ugent.be" target="_blank">michael.sluydts@ugent.be</a>&gt;</span><br>
<blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
  
    
  
  <div text="#000000" bgcolor="#FFFFFF">
    <div>While I&#39;m not sure how easily normal
      desktop benchmarks transfer to parallel processing through wien2k
      I usually look at the following benchmarks when comparing CPUs
      (and prices): <br>
      <br>
      <a href="http://cpubenchmark.net/high_end_cpus.html" target="_blank">http://cpubenchmark.net/high_end_cpus.html</a><br>
      <br>
      <br>
      Best regards,<br>
      <br>
      Michael Sluydts<br>
      <br>
      Op 12/09/2013 15:43, Luis Ogando schreef:<br>
    </div><div><div class="h5">
    <blockquote type="cite">
      <div dir="ltr">Dear Prof. Blaha,
        <div><br>
        </div>
        <div>   Thank you very much for the explanations. They
          will be very useful !!</div>
        <div>   All the best,</div>
        <div>                        Luis</div>
      </div>
      <div class="gmail_extra">
        <br>
        <br>
        <div class="gmail_quote">2013/9/12 Peter Blaha <span dir="ltr">&lt;<a href="mailto:pblaha@theochem.tuwien.ac.at" target="_blank">pblaha@theochem.tuwien.ac.at</a>&gt;</span><br>
          <blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
            This depends a lot on what you want to do and how much money
            you have.<br>
            <br>
            The single-core speed of a fast I7 is at least as good (or
            faster) than most Xeons, and they are MUCH cheaper. So for
            all systems up to 64-100 atoms/cell, where you need several
            k-points, a small cluster of I7 cpus<br>
            /4 cores, or more expensive 6 core) (with GB-network and a
            common NFS file system) is for sure the fastest platform and
            in particular has by FAR the best price/performance ratio
            (One powerful I7 computer may cost about 1000 Euro). For
            bigger clusters, a drawback can be the large &quot;space&quot; to put
            all PCs on a big shelf ....), but if you have less than
            10000 Euros, this is probably the best choice.<br>
            <br>
            However, Xeons can be coupled (2-4 Xeons) to a &quot;single
            multicore computer (eg. 16 cores)&quot;, which may work with mpi
            and can be used to handle systems up to 200-300 atoms. They
            also can be bought in small boxes and may fit in a single 19
            inch cabinet. But of course such systems are much more
            expensive. From what I said above it should be clear, that
            it is completely useless to buy a &quot;single 4 core Xeon
            computer&quot;.<br>
            <br>
            The next step would be to buy an Infiniband switch+cards and
            couple your PCs with this fast network to a powerful
            multinode mpi-cluster. Since the switch/cards are fairly
            expensive, on usually takes here Xeons as platform. However,
            you need to know how to install/configure the software
            properly. I&#39;ve seen such clusters even in computing centers,
            which were completely useless, because the network/mpi was
            instable and jobs would crash randomly every couple of hours
            .....<br>
            <br>
            Our strategy:<br>
            i) We have a GB-networked cluster with Intel I7 computers
            (which we maintain our-self and this cluster includes also
            all the user- workstations) and do all the calculations for
            systems up to 64 atoms/cell on these systems.<br>
            2) For bigger systems we go to our University
            computer-center and run there with a PBS queuing system.
            This has the advantage that we do not need to care about the
            installation of the infiniband network nor the
            mpi-infrastructure (but we use always intel-mpi together
            with ifort/mkl).
            <div>
              <br>
              <br>
              On 09/11/2013 06:31 PM, Luis Ogando wrote:<br>
            </div>
            <blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
              <div>
                Dear Prof. Blaha,<br>
                <br>
                    Just for curiosity, what processor did you buy ?<br>
                    Is the Xeon family better than the i7 one for WIEN2k
                calculations ?<br>
                    All the best,<br>
                                        Luis<br>
                <br>
                <br>
                2013/9/11 Peter Blaha &lt;<a href="mailto:pblaha@theochem.tuwien.ac.at" target="_blank">pblaha@theochem.tuwien.ac.at</a><br>
              </div>
              &lt;mailto:<a href="mailto:pblaha@theochem.tuwien.ac.at" target="_blank">pblaha@theochem.tuwien.ac.at</a>&gt;&gt;
              <div><br>
                <br>
                    I don&#39;t know what &quot;latest&quot; means. We use the latest
                one installed on<br>
                    our supercomputers (4.1.1.036)<br>
                <br>
                    I have not seen any significant change with mpi in
                the last years.<br>
                <br>
                    PS: I just got info that we have now a new ifort
                available for<br>
                    download ...<br>
                <br>
                <br>
                    On 09/11/2013 05:00 PM, Laurence Marks wrote:<br>
                <br>
                        Thanks.<br>
                <br>
                        One thing I will add/ask concerning the
                parallelization, the latest<br>
                        impi seems to be substantially better -- have
                you tried it? I<br>
                        have not<br>
                        just noticed this with Wien2k, but I am told
                that others have seen<br>
                        improvements in other codes.<br>
                <br>
                        On Wed, Sep 11, 2013 at 9:42 AM, Peter Blaha<br>
                        &lt;<a href="mailto:pblaha@theochem.tuwien.ac.at" target="_blank">pblaha@theochem.tuwien.ac.at</a><br>
              </div>
              <div>
                <div>
                          &lt;mailto:<a href="mailto:pblaha@theochem.tuwien.ac.at" target="_blank">pblaha@theochem.tuwien.ac.at</a>&gt;&gt;
                  wrote:<br>
                  <br>
                              Before buying a couple of new computers, I
                  was asking myself<br>
                              the same<br>
                              question and discussed this with some
                  people of our<br>
                              computing departments.<br>
                  <br>
                              The conclusions:<br>
                              a) potentially very good, but in practice
                  very questionable,<br>
                              because for<br>
                              most application you cannot get out the
                  real speed (10 times<br>
                              faster than<br>
                              an Intel I7). This is true, even for many
                  lapack/mkl<br>
                              subroutines where<br>
                              it &quot;should&quot; work better.<br>
                              They told me to &quot;wait&quot;, until the mkl
                  becomes better<br>
                              (hopefully). I&#39;m<br>
                              not too optimistic, when you see how badly
                  the<br>
                              mkl-parallelization of<br>
                              multicore machines is working (2 cores is
                  very good, but 4<br>
                              or more is<br>
                              already very bad).<br>
                  <br>
                              b) The nature of our problem (big
                  eigenvalue problem): A &quot;fast<br>
                              processor&quot; is useful only for large
                  problems --&gt; large memory.<br>
                              You can buy Phi coprocessors now with
                  quite some large<br>
                              memory, but then<br>
                              they are terrible expensive (and 5
                  &quot;normal&quot; PCs are faster<br>
                              and cheaper)<br>
                  <br>
                              c) the hardware design has a VERY slow
                  communication between<br>
                              main-memory<br>
                              and Phi-memory. This makes also
                  parallelization over several<br>
                              PHI-nodes<br>
                              via mpi not really possible (if you need
                  any significant<br>
                              data transfer,<br>
                              like for an eigenvalue problem).<br>
                  <br>
                              Thus I did not buy it.<br>
                  <br>
                              However, if anybody has access and time to
                  try out WIEN2k on<br>
                              PHis, I&#39;d<br>
                              would be very interested in getting
                  feedback. (Maybe these<br>
                              computer-people were not good enough ....)<br>
                  <br>
                              PS: I know from G.Kresse that they had
                  some time ago (maybe<br>
                              2 years ?)<br>
                              an expert from NVIDIA with them. After 2
                  weeks of porting<br>
                              VASP to these<br>
                              GPUs by this expert, VASP on the GPU was
                  &quot;almost as fast&quot; as<br>
                              on an Intel<br>
                              I7 processor.<br>
                  <br>
                  <br>
                              On 09/11/2013 04:16 PM, Laurence Marks
                  wrote:<br>
                  <br>
                                  Anyone know if these will be viable
                  with Wien2k (mpi,<br>
                                  i.e. large problems)?<br>
                  <br>
                  <br>
                              --<br>
                  <br>
                                                                     
                   P.Blaha<br>
                </div>
              </div>
                          ------------------------------__------------------------------__--------------
              <div><br>
                            Peter BLAHA, Inst.f. Materials Chemistry, TU
                Vienna, A-1060<br>
                            Vienna<br>
              </div>
                          Phone: <a href="tel:%2B43-1-58801-165300" value="+43158801165300" target="_blank">+43-1-58801-165300</a>
              <a href="tel:%2B43-1-58801-165300" target="_blank">&lt;tel:%2B43-1-58801-165300&gt;</a><br>
                               FAX: <a href="tel:%2B43-1-58801-165982" value="+43158801165982" target="_blank">+43-1-58801-165982</a>
              <a href="tel:%2B43-1-58801-165982" target="_blank">&lt;tel:%2B43-1-58801-165982&gt;</a><br>
                          Email: <a href="mailto:blaha@theochem.tuwien.ac.at" target="_blank">blaha@theochem.tuwien.ac.at</a><br>
                          &lt;mailto:<a href="mailto:blaha@theochem.tuwien.ac.at" target="_blank">blaha@theochem.tuwien.ac.at</a>&gt;  
               WWW:<br>
                          <a href="http://info.tuwien.ac.at/__theochem/" target="_blank">http://info.tuwien.ac.at/__theochem/</a><br>
                          &lt;<a href="http://info.tuwien.ac.at/theochem/" target="_blank">http://info.tuwien.ac.at/theochem/</a>&gt;<br>
                          ------------------------------__------------------------------__--------------<br>
                          _________________________________________________<br>
                          Wien mailing list<br>
                          <a href="mailto:Wien@zeus.theochem.tuwien.ac.__at" target="_blank">Wien@zeus.theochem.tuwien.ac.__at</a><br>
                          &lt;mailto:<a href="mailto:Wien@zeus.theochem.tuwien.ac.at" target="_blank">Wien@zeus.theochem.tuwien.ac.at</a>&gt;<br>
                          <a href="http://zeus.theochem.tuwien." target="_blank">http://zeus.theochem.tuwien.</a>__<a href="http://ac.at/mailman/listinfo/wien" target="_blank">ac.at/mailman/listinfo/wien</a>
              <div><br>
                            &lt;<a href="http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien" target="_blank">http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien</a>&gt;<br>
                            SEARCH the MAILING-LIST at:<br>
              </div>
                          <a href="http://www.mail-archive.com/__wien@zeus.theochem.tuwien.ac.__at/index.html" target="_blank">http://www.mail-archive.com/__wien@zeus.theochem.tuwien.ac.__at/index.html</a><br>
                          &lt;<a href="http://www.mail-archive.com/wien@zeus.theochem.tuwien.ac.at/index.html" target="_blank">http://www.mail-archive.com/wien@zeus.theochem.tuwien.ac.at/index.html</a>&gt;<br>
              <br>
              <br>
              <br>
              <br>
              <br>
                  --<br>
              <br>
                                                         P.Blaha<br>
                  ------------------------------__------------------------------__--------------
              <div><br>
                    Peter BLAHA, Inst.f. Materials Chemistry, TU Vienna,
                A-1060 Vienna<br>
              </div>
                  Phone: <a href="tel:%2B43-1-58801-165300" value="+43158801165300" target="_blank">+43-1-58801-165300</a>
              <a href="tel:%2B43-1-58801-165300" target="_blank">&lt;tel:%2B43-1-58801-165300&gt;</a><br>
                  FAX: <a href="tel:%2B43-1-58801-165982" value="+43158801165982" target="_blank">+43-1-58801-165982</a>
              <a href="tel:%2B43-1-58801-165982" target="_blank">&lt;tel:%2B43-1-58801-165982&gt;</a><br>
                  Email: <a href="mailto:blaha@theochem.tuwien.ac.at" target="_blank">blaha@theochem.tuwien.ac.at</a><br>
                  &lt;mailto:<a href="mailto:blaha@theochem.tuwien.ac.at" target="_blank">blaha@theochem.tuwien.ac.at</a>&gt;  
               WWW:<br>
                  <a href="http://info.tuwien.ac.at/__theochem/" target="_blank">http://info.tuwien.ac.at/__theochem/</a><br>
                  &lt;<a href="http://info.tuwien.ac.at/theochem/" target="_blank">http://info.tuwien.ac.at/theochem/</a>&gt;<br>
                  ------------------------------__------------------------------__--------------<br>
                  _________________________________________________<br>
                  Wien mailing list<br>
                  <a href="mailto:Wien@zeus.theochem.tuwien.ac.__at" target="_blank">Wien@zeus.theochem.tuwien.ac.__at</a><br>
                  &lt;mailto:<a href="mailto:Wien@zeus.theochem.tuwien.ac.at" target="_blank">Wien@zeus.theochem.tuwien.ac.at</a>&gt;<br>
                  <a href="http://zeus.theochem.tuwien." target="_blank">http://zeus.theochem.tuwien.</a>__<a href="http://ac.at/mailman/listinfo/wien" target="_blank">ac.at/mailman/listinfo/wien</a>
              <div><br>
                    &lt;<a href="http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien" target="_blank">http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien</a>&gt;<br>
                    SEARCH the MAILING-LIST at:<br>
              </div>
                  <a href="http://www.mail-archive.com/__wien@zeus.theochem.tuwien.ac.__at/index.html" target="_blank">http://www.mail-archive.com/__wien@zeus.theochem.tuwien.ac.__at/index.html</a><br>
                  &lt;<a href="http://www.mail-archive.com/wien@zeus.theochem.tuwien.ac.at/index.html" target="_blank">http://www.mail-archive.com/wien@zeus.theochem.tuwien.ac.at/index.html</a>&gt;
              <div><br>
                <br>
                <br>
                <br>
                <br>
                _______________________________________________<br>
                Wien mailing list<br>
                <a href="mailto:Wien@zeus.theochem.tuwien.ac.at" target="_blank">Wien@zeus.theochem.tuwien.ac.at</a><br>
                <a href="http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien" target="_blank">http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien</a><br>
                SEARCH the MAILING-LIST at:  <a href="http://www.mail-archive.com/wien@zeus.theochem.tuwien.ac.at/index.html" target="_blank">http://www.mail-archive.com/wien@zeus.theochem.tuwien.ac.at/index.html</a><br>

                <br>
              </div>
            </blockquote>
            <div>
              <div>
                <br>
                -- <br>
                <br>
                                                      P.Blaha<br>
                --------------------------------------------------------------------------<br>
                Peter BLAHA, Inst.f. Materials Chemistry, TU Vienna,
                A-1060 Vienna<br>
                Phone: <a href="tel:%2B43-1-58801-165300" value="+43158801165300" target="_blank">+43-1-58801-165300</a>
                            FAX: <a href="tel:%2B43-1-58801-165982" value="+43158801165982" target="_blank">+43-1-58801-165982</a><br>
                Email: <a href="mailto:blaha@theochem.tuwien.ac.at" target="_blank">blaha@theochem.tuwien.ac.at</a>  
                 WWW: <a href="http://info.tuwien.ac.at/theochem/" target="_blank">http://info.tuwien.ac.at/theochem/</a><br>
                --------------------------------------------------------------------------<br>
                _______________________________________________<br>
                Wien mailing list<br>
                <a href="mailto:Wien@zeus.theochem.tuwien.ac.at" target="_blank">Wien@zeus.theochem.tuwien.ac.at</a><br>
                <a href="http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien" target="_blank">http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien</a><br>
                SEARCH the MAILING-LIST at:  <a href="http://www.mail-archive.com/wien@zeus.theochem.tuwien.ac.at/index.html" target="_blank">http://www.mail-archive.com/wien@zeus.theochem.tuwien.ac.at/index.html</a><br>

              </div>
            </div>
          </blockquote>
        </div>
        <br>
      </div>
      <br>
      <fieldset></fieldset>
      <br>
      <pre>_______________________________________________
Wien mailing list
<a href="mailto:Wien@zeus.theochem.tuwien.ac.at" target="_blank">Wien@zeus.theochem.tuwien.ac.at</a>
<a href="http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien" target="_blank">http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien</a>
SEARCH the MAILING-LIST at:  <a href="http://www.mail-archive.com/wien@zeus.theochem.tuwien.ac.at/index.html" target="_blank">http://www.mail-archive.com/wien@zeus.theochem.tuwien.ac.at/index.html</a>
</pre>
    </blockquote>
    <br>
  </div></div></div>

<br>_______________________________________________<br>
Wien mailing list<br>
<a href="mailto:Wien@zeus.theochem.tuwien.ac.at">Wien@zeus.theochem.tuwien.ac.at</a><br>
<a href="http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien" target="_blank">http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien</a><br>
SEARCH the MAILING-LIST at:  <a href="http://www.mail-archive.com/wien@zeus.theochem.tuwien.ac.at/index.html" target="_blank">http://www.mail-archive.com/wien@zeus.theochem.tuwien.ac.at/index.html</a><br>
<br></blockquote></div><br></div>