<?xml version="1.0" encoding="UTF-8"?>
<rss xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:rdf="http://www.w3.org/1999/02/22-rdf-syntax-ns#" xmlns:taxo="http://purl.org/rss/1.0/modules/taxonomy/" version="2.0">
  <channel>
    <title>topic Re: Performance Counters and IMC Counter Not Matching in Software Tuning, Performance Optimization &amp; Platform Monitoring</title>
    <link>https://community.intel.com/t5/Software-Tuning-Performance/Performance-Counters-and-IMC-Counter-Not-Matching/m-p/1293267#M7901</link>
    <description>&lt;P&gt;Thank you John,&lt;/P&gt;
&lt;P&gt;There was an &lt;STRONG&gt;error&lt;/STRONG&gt; in &lt;STRONG&gt;IMC counter&lt;/STRONG&gt; value which is &lt;STRONG&gt;revised&lt;/STRONG&gt; now. It was &lt;STRONG&gt;ten&lt;/STRONG&gt; times greater than the &lt;STRONG&gt;real&lt;/STRONG&gt; value. Now it is `100` times the value of &lt;STRONG&gt;CPU counters&lt;/STRONG&gt;.&amp;nbsp;&lt;/P&gt;
&lt;P&gt;I wonder why the &lt;STRONG&gt;two&lt;/STRONG&gt; counters do &lt;STRONG&gt;not&lt;/STRONG&gt; match. I mean if &lt;STRONG&gt;all&lt;/STRONG&gt; (or almost all) of the accesses &lt;STRONG&gt;originate&lt;/STRONG&gt; from the &lt;STRONG&gt;cores&lt;/STRONG&gt; (&lt;STRONG&gt;not&lt;/STRONG&gt; sth such as &lt;STRONG&gt;IO&lt;/STRONG&gt; devices), they should &lt;STRONG&gt;show&lt;/STRONG&gt; their effect in &lt;STRONG&gt;both&lt;/STRONG&gt; counters (&lt;STRONG&gt;not&lt;/STRONG&gt; just the &lt;STRONG&gt;IMC&lt;/STRONG&gt; counter).&lt;/P&gt;
&lt;P&gt;Regards.&lt;/P&gt;</description>
    <pubDate>Fri, 25 Jun 2021 12:35:23 GMT</pubDate>
    <dc:creator>TheAhmad</dc:creator>
    <dc:date>2021-06-25T12:35:23Z</dc:date>
    <item>
      <title>Performance Counters and IMC Counter Not Matching</title>
      <link>https://community.intel.com/t5/Software-Tuning-Performance/Performance-Counters-and-IMC-Counter-Not-Matching/m-p/1288832#M7886</link>
      <description>&lt;P&gt;Hi,&lt;/P&gt;
&lt;P&gt;I ran `Perf` on a relatively &lt;STRONG&gt;idle&lt;/STRONG&gt; system to &lt;STRONG&gt;count&lt;/STRONG&gt; the number of &lt;STRONG&gt;DRAM accesses&lt;/STRONG&gt;. I also used the &lt;EM&gt;internal&lt;/EM&gt; `IMC` counter to count accesses. But the value obtained using the &lt;STRONG&gt;former method&lt;/STRONG&gt; is &lt;STRONG&gt;much much smaller&lt;/STRONG&gt; than the &lt;STRONG&gt;latter&lt;/STRONG&gt;. Does anybody have &lt;STRONG&gt;ANY&lt;/STRONG&gt; guesses about the reason?&lt;/P&gt;
&lt;P&gt;The link to the &lt;STRONG&gt;detailed&lt;/STRONG&gt; question is followed:&lt;/P&gt;
&lt;P&gt;&lt;A href="https://stackoverflow.com/questions/67513607/performance-counters-and-imc-counter-not-matching" target="_blank"&gt;https://stackoverflow.com/questions/67513607/performance-counters-and-imc-counter-not-matching&lt;/A&gt;&lt;/P&gt;
&lt;P&gt;Regards.&lt;/P&gt;</description>
      <pubDate>Thu, 10 Jun 2021 17:01:45 GMT</pubDate>
      <guid>https://community.intel.com/t5/Software-Tuning-Performance/Performance-Counters-and-IMC-Counter-Not-Matching/m-p/1288832#M7886</guid>
      <dc:creator>TheAhmad</dc:creator>
      <dc:date>2021-06-10T17:01:45Z</dc:date>
    </item>
    <item>
      <title>Re: Performance Counters and IMC Counter Not Matching</title>
      <link>https://community.intel.com/t5/Software-Tuning-Performance/Performance-Counters-and-IMC-Counter-Not-Matching/m-p/1290418#M7893</link>
      <description>&lt;P&gt;Uncontrolled measurements over different time intervals are unlikely to match.&lt;/P&gt;
&lt;P&gt;Your L3 miss rate of 300,000 per second is tiny and could have a large number of different sources.&lt;/P&gt;
&lt;P&gt;The UNC_IMC_DRAM_DATA_READS counter is the sum of DRAM CAS Reads on the two channels. &amp;nbsp;A value of 311 million per second corresponds to 19.9 GB/s -- about 78% of the peak DRAM bandwidth. &amp;nbsp; &amp;nbsp;This seems too large to be solely due to IO.&lt;/P&gt;
&lt;P&gt;The behavior of "idle" systems is increasingly difficult to understand as the operating system sees an "idle" system as an opportunity to do as much of its own maintenance work as possible without interfering with user processes. &amp;nbsp; Detailed tracing of activity may explain what is happening, but it may also take months to learn enough about how the kernel works to understand the measurements.&lt;/P&gt;</description>
      <pubDate>Wed, 16 Jun 2021 16:18:37 GMT</pubDate>
      <guid>https://community.intel.com/t5/Software-Tuning-Performance/Performance-Counters-and-IMC-Counter-Not-Matching/m-p/1290418#M7893</guid>
      <dc:creator>McCalpinJohn</dc:creator>
      <dc:date>2021-06-16T16:18:37Z</dc:date>
    </item>
    <item>
      <title>Re: Performance Counters and IMC Counter Not Matching</title>
      <link>https://community.intel.com/t5/Software-Tuning-Performance/Performance-Counters-and-IMC-Counter-Not-Matching/m-p/1293267#M7901</link>
      <description>&lt;P&gt;Thank you John,&lt;/P&gt;
&lt;P&gt;There was an &lt;STRONG&gt;error&lt;/STRONG&gt; in &lt;STRONG&gt;IMC counter&lt;/STRONG&gt; value which is &lt;STRONG&gt;revised&lt;/STRONG&gt; now. It was &lt;STRONG&gt;ten&lt;/STRONG&gt; times greater than the &lt;STRONG&gt;real&lt;/STRONG&gt; value. Now it is `100` times the value of &lt;STRONG&gt;CPU counters&lt;/STRONG&gt;.&amp;nbsp;&lt;/P&gt;
&lt;P&gt;I wonder why the &lt;STRONG&gt;two&lt;/STRONG&gt; counters do &lt;STRONG&gt;not&lt;/STRONG&gt; match. I mean if &lt;STRONG&gt;all&lt;/STRONG&gt; (or almost all) of the accesses &lt;STRONG&gt;originate&lt;/STRONG&gt; from the &lt;STRONG&gt;cores&lt;/STRONG&gt; (&lt;STRONG&gt;not&lt;/STRONG&gt; sth such as &lt;STRONG&gt;IO&lt;/STRONG&gt; devices), they should &lt;STRONG&gt;show&lt;/STRONG&gt; their effect in &lt;STRONG&gt;both&lt;/STRONG&gt; counters (&lt;STRONG&gt;not&lt;/STRONG&gt; just the &lt;STRONG&gt;IMC&lt;/STRONG&gt; counter).&lt;/P&gt;
&lt;P&gt;Regards.&lt;/P&gt;</description>
      <pubDate>Fri, 25 Jun 2021 12:35:23 GMT</pubDate>
      <guid>https://community.intel.com/t5/Software-Tuning-Performance/Performance-Counters-and-IMC-Counter-Not-Matching/m-p/1293267#M7901</guid>
      <dc:creator>TheAhmad</dc:creator>
      <dc:date>2021-06-25T12:35:23Z</dc:date>
    </item>
    <item>
      <title>Re: Performance Counters and IMC Counter Not Matching</title>
      <link>https://community.intel.com/t5/Software-Tuning-Performance/Performance-Counters-and-IMC-Counter-Not-Matching/m-p/1311321#M7932</link>
      <description>&lt;P&gt;I have found the answer. The question is answered at:&lt;/P&gt;
&lt;P&gt;&lt;A href="https://stackoverflow.com/a/69006281/6661026" target="_self"&gt;Stackoverflow&lt;/A&gt;&amp;nbsp;.&lt;/P&gt;</description>
      <pubDate>Wed, 01 Sep 2021 05:52:57 GMT</pubDate>
      <guid>https://community.intel.com/t5/Software-Tuning-Performance/Performance-Counters-and-IMC-Counter-Not-Matching/m-p/1311321#M7932</guid>
      <dc:creator>TheAhmad</dc:creator>
      <dc:date>2021-09-01T05:52:57Z</dc:date>
    </item>
  </channel>
</rss>

