<div>Hello Petsc team (especially Satish and Barry).</div>
<div> </div>
<div>YOU SAID: FOR Better performance </div>
<div> </div>
<div>(1) <span style="FONT-WEIGHT: bold">high per-CPU memory performance</span>. Each CPU (core in dual core systems) needs to have its <span style="FONT-WEIGHT: bold">own</span> memory bandwith of roughly 2 or more gigabytes.</div>
<div> </div>
<div>(2) MEMORY BANDWDITH PER CORE, the higher that is the better performance you get.<br> </div>
<div>From these points I started to look for RAM Sticks with higher MHz rates (and obviously CPUs and motherboards supporting this speed). </div>
<div> </div>
<div>But you also reflected to:</div>
<div><br><a onclick="return top.js.OpenExtLink(window,event,this)" href="http://www.intel.com/performance/server/xeon/hpc_ansys.htm" target="_blank">http://www.intel.com/performance/server/xeon/hpc_ansys.htm</a><br><a onclick="return top.js.OpenExtLink(window,event,this)" href="http://www.intel.com/performance/server/xeon/hpc_md_nastr.htm" target="_blank">http://www.intel.com/performance/server/xeon/hpc_md_nastr.htm</a><br>
</div>
<div>On these pages you pointed out that: systems with CPUs of 20% higher FSB speed are performing 20% better. But you see also RAM speed is 20% higher for the better performing system (i.e 800MHz vs 667 MHz).</div>
<div> </div>
<div>So my question is that which is the actual indicator of "memory bandwidth"per core?</div>
<div>Whether it is </div>
<div>(1) CPU's FSB speed </div>
<div>(2) RAM speed </div>
<div>(3) Motherboard's System Bus Speed.</div>
<div> </div>
<div>How we could ensure "memory bandwith of roughly 2 or more gigabytes" per CPU core? (Higher CPU's FSB speed, or RAM speed or Motherboard's System Bus Speed).</div>
<div> </div>
<div>With best regards,</div>
<div>Amjad Ali.</div>
<div> </div>
<div> </div>
<div> </div>
<div><br><br> </div>
<div><span class="gmail_quote">On 4/16/08, <b class="gmail_sendername">Barry Smith</b> <<a href="mailto:bsmith@mcs.anl.gov">bsmith@mcs.anl.gov</a>> wrote:</span>
<blockquote class="gmail_quote" style="PADDING-LEFT: 1ex; MARGIN: 0px 0px 0px 0.8ex; BORDER-LEFT: #ccc 1px solid"><br> Cool. The pages to look at are<br><br><a onclick="return top.js.OpenExtLink(window,event,this)" href="http://www.intel.com/performance/server/xeon/hpc_ansys.htm" target="_blank">http://www.intel.com/performance/server/xeon/hpc_ansys.htm</a><br>
<a onclick="return top.js.OpenExtLink(window,event,this)" href="http://www.intel.com/performance/server/xeon/hpc_md_nastr.htm" target="_blank">http://www.intel.com/performance/server/xeon/hpc_md_nastr.htm</a><br><br>these are the two benchmarks that reflect the bottlenecks of memory bandwidth.<br>
When going from dual to quad they get 1.2 times the performance, when one would<br>like 2 times the performance.<br><span class="sg"><br> Barry</span>
<div><span class="e" id="q_11957c014d7627ec_2"><br><br><br>On Apr 16, 2008, at 9:27 AM, Satish Balay wrote:<br>
<blockquote class="gmail_quote" style="PADDING-LEFT: 1ex; MARGIN: 0px; BORDER-LEFT: #ccc 1px solid">Just a note:<br><br>Intel does publish benchmarks for their chips.<br><br><a onclick="return top.js.OpenExtLink(window,event,this)" href="http://www.intel.com/performance/server/xeon/hpcapp.htm" target="_blank">http://www.intel.com/performance/server/xeon/hpcapp.htm</a><br>
<br>Satish<br><br></blockquote><br></span></div></blockquote></div><br>