<html><body>
<DIV><BR>"andrew holway" <andrew@moonet.co.uk> wrote:</DIV>
<DIV><BR>> <BR>> I am looking at chips for a small 20 node cluster for computational <BR>> chemistry. I'm trying to get my head around the various merits of the <BR>> different ways the cache work. <BR>> <BR>> Considering the fairly chunky bits of data that are going to be going <BR>> through, does anyone have any opinion on how the new 1.9 Barcelona <BR>> will perform against a 2.66 woodcrest. Is the Intel front side bus <BR>> going to be a significant bottleneck for this type of application?</DIV>
<DIV> </DIV>
<DIV>Andrew,</DIV>
<DIV> </DIV>
<DIV>You should be comparing Barcelona to Harpertown, and focused</DIV>
<DIV>on what things will look like in Q4 when Harpertown (45nm) will be out</DIV>
<DIV>and Barcelona "will" be available at higher clock. Barcelona is still the </DIV>
<DIV>bandwidth and latency leader, and compares close to equal on power</DIV>
<DIV>with the Harpertown at 45nm. Codes that favor larger caches will</DIV>
<DIV>like Harpertown. If the Barcelona makes it to 2.6 or 2.8 GHz by </DIV>
<DIV>the end of the year, then it will compete. It will win with even at lower </DIV>
<DIV>clocks on very bandwidth intensive apps.</DIV>
<DIV> </DIV>
<DIV>Regards,</DIV>
<DIV> </DIV>
<DIV>rbw</DIV>
<DIV><BR>-- <BR><BR>"Making predictions is hard, especially about the future." <BR><BR>Niels Bohr <BR><BR>-- <BR><BR>Richard Walsh <BR>Thrashing River Consulting-- <BR>5605 Alameda St. <BR>Shoreview, MN 55126 <BR></DIV></body></html>