<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN">
<html><head><title>Re[2]: [Beowulf] Nehalem memory configs</title>
<META http-equiv=Content-Type content="text/html; charset=iso-8859-1">
<meta http-equiv="Content-Style-Type" content="text/css">
<style type="text/css"><!--
body {
margin: 5px 5px 5px 5px;
background-color: #ffffff;
}
/* ========== Text Styles ========== */
hr { color: #000000}
body, table /* Normal text */
{
font-size: 9pt;
font-family: 'Courier New';
font-style: normal;
font-weight: normal;
color: #000000;
text-decoration: none;
}
span.rvts1 /* Heading */
{
font-size: 10pt;
font-family: 'Arial';
font-weight: bold;
color: #0000ff;
}
span.rvts2 /* Subheading */
{
font-size: 10pt;
font-family: 'Arial';
font-weight: bold;
color: #000080;
}
span.rvts3 /* Keywords */
{
font-size: 10pt;
font-family: 'Arial';
font-style: italic;
color: #800000;
}
a.rvts4, span.rvts4 /* Jump 1 */
{
font-size: 10pt;
font-family: 'Arial';
color: #008000;
text-decoration: underline;
}
a.rvts5, span.rvts5 /* Jump 2 */
{
font-size: 10pt;
font-family: 'Arial';
color: #008000;
text-decoration: underline;
}
span.rvts6
{
font-weight: bold;
color: #800000;
}
span.rvts7
{
font-weight: bold;
color: #800080;
}
a.rvts8, span.rvts8
{
color: #0000ff;
text-decoration: underline;
}
span.rvts9
{
font-weight: bold;
color: #800000;
}
/* ========== Para Styles ========== */
p,ul,ol /* Paragraph Style */
{
text-align: left;
text-indent: 0px;
padding: 0px 0px 0px 0px;
margin: 0px 0px 0px 0px;
}
.rvps1 /* Centered */
{
text-align: center;
}
--></style>
</head>
<body>
<p>Hallo Jeff,</p>
<p><br></p>
<p>Mittwoch, 8. April 2009, meintest Du:</p>
<p><br></p>
<p><span class=rvts6>JL> Peter Kjellstrom wrote:</span></p>
<p><span class=rvts7>>> On Wednesday 08 April 2009, Joe Landman wrote:</span></p>
<p><span class=rvts7>>> </span></p>
<p><span class=rvts6>>>> As an FYI, Beowulf veteran Jeff Layton wrote up a nice article on memory</span></p>
<p><span class=rvts6>>>> configuration issues for Nehalem (I had seen some discussion on this</span></p>
<p><span class=rvts6>>>> previously).</span></p>
<p><span class=rvts6>>>> </span></p>
<p><br></p>
<p><span class=rvts7>>> Can anyone confirm that mixing different sizes of dimms, even when keeping to </span></p>
<p><span class=rvts7>>> a by-three-symmetric configuration, actually does degrade performance?</span></p>
<p><br></p>
<p><span class=rvts7>>> That is, first hand information that config 1 will be slower than config 2:</span></p>
<p><br></p>
<p><span class=rvts7>>> 1: two sockets, 3x 1G and 3x 2G per socket RDIMM (total 18G @1066)</span></p>
<p><span class=rvts7>>> 2: two sockets, 6x 1G(or 2G) per socket RDIMM (total 12G (or 24G) @1066)</span></p>
<p><br></p>
<p><span class=rvts7>>> /Peter</span></p>
<p><span class=rvts7>>> </span></p>
<p><br></p>
<p><span class=rvts6>JL> I can confirm it because I've got it on a test box I've been running :)</span></p>
<p><br></p>
<p>Can you give us the result of a stream-benchmark? with 12 x RDIMM in asymmetric config and 6 x RDIMM ? both running at 1066?</p>
<p><br></p>
<p>Would be great!</p>
<p><br></p>
<p>Jan</p>
</body></html>