<?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 Max. IOH speed, max GPUs speed in Graphics</title>
    <link>https://community.intel.com/t5/Graphics/Max-IOH-speed-max-GPUs-speed/m-p/210487#M1710</link>
    <description>&lt;P&gt;There are two questions really.&lt;/P&gt;&lt;P&gt;&lt;/P&gt;&lt;P&gt;The first is: what is the  5520 chipset  to processor real max. speed?&lt;/P&gt;&lt;P&gt;Is it limited by the QPI single-link or both links may work together? (in other words, is it limited to  raw 12.8GB/s or may support all four PCIe links at a time?&lt;/P&gt;&lt;P&gt;&lt;/P&gt;&lt;P&gt;The second is: have Intel mesured real bandwidth used by most modern AMD/NVIDIA GPU/GPGPUs?&lt;/P&gt;&lt;P&gt;It's clear by their internal architecture that their theoretical max throughput may not exeed 1.25GB/s (so up to 6 GPUs may nicely share one x16 link (providing them all by switched x16 links). But is it possible to share the link between, say 8 (or even more) GPUs?&lt;/P&gt;&lt;P&gt;&lt;/P&gt;&lt;P&gt;The number of PCIe lines is limited, but a fart workstation needs not only ultra-fast graphics, but ultra-fast storage too.&lt;/P&gt;</description>
    <pubDate>Sat, 04 Sep 2010 13:35:33 GMT</pubDate>
    <dc:creator>idata</dc:creator>
    <dc:date>2010-09-04T13:35:33Z</dc:date>
    <item>
      <title>Max. IOH speed, max GPUs speed</title>
      <link>https://community.intel.com/t5/Graphics/Max-IOH-speed-max-GPUs-speed/m-p/210487#M1710</link>
      <description>&lt;P&gt;There are two questions really.&lt;/P&gt;&lt;P&gt;&lt;/P&gt;&lt;P&gt;The first is: what is the  5520 chipset  to processor real max. speed?&lt;/P&gt;&lt;P&gt;Is it limited by the QPI single-link or both links may work together? (in other words, is it limited to  raw 12.8GB/s or may support all four PCIe links at a time?&lt;/P&gt;&lt;P&gt;&lt;/P&gt;&lt;P&gt;The second is: have Intel mesured real bandwidth used by most modern AMD/NVIDIA GPU/GPGPUs?&lt;/P&gt;&lt;P&gt;It's clear by their internal architecture that their theoretical max throughput may not exeed 1.25GB/s (so up to 6 GPUs may nicely share one x16 link (providing them all by switched x16 links). But is it possible to share the link between, say 8 (or even more) GPUs?&lt;/P&gt;&lt;P&gt;&lt;/P&gt;&lt;P&gt;The number of PCIe lines is limited, but a fart workstation needs not only ultra-fast graphics, but ultra-fast storage too.&lt;/P&gt;</description>
      <pubDate>Sat, 04 Sep 2010 13:35:33 GMT</pubDate>
      <guid>https://community.intel.com/t5/Graphics/Max-IOH-speed-max-GPUs-speed/m-p/210487#M1710</guid>
      <dc:creator>idata</dc:creator>
      <dc:date>2010-09-04T13:35:33Z</dc:date>
    </item>
  </channel>
</rss>

