Author: Robert Hyatt
Date: 22:30:18 05/15/04
Go up one level in this thread
On May 14, 2004 at 20:43:24, Vincent Diepeveen wrote: >On May 14, 2004 at 08:23:41, Robert Hyatt wrote: > >The difference between some cc-NUMA machines and clusters is only the fact that >the first one uses software from SGI and the second one from IBM. > >It's just a software layer. Utter and total nonsense. But I won't bore others with the datails of why that is so far beyond ridiculous that it doesn't even deserve a response. but NUMA --> shared memory. Clusters don't do shared memory. > >>On May 13, 2004 at 23:58:43, Vincent Diepeveen wrote: >> >>>On May 13, 2004 at 12:46:29, Robert Hyatt wrote: >>> >>>>On May 13, 2004 at 07:47:32, Vincent Diepeveen wrote: >>>> >>>>>On May 13, 2004 at 03:09:22, Joshua Shriver wrote: >>>>> >>>>>>Question is... are they going to run a parallel based chess engine on it :) >>>>> >>>>>Only diep would run at it, and they didn't approach me. >>>> >>>>I can't _possibly_ imagine why "they don't approach you." >>>> >>>>Also you do know this is a cluster? Not a NUMA box? >>>> >>>>Of course you did. >>>> >>>>I'm sure your program does _great_ on a message-passing cluster... >>> >>>May i remind you your message passing cluster you get there has a faster one way >>>pingpong latency than origin3800 at 512 processors.... >> >>May I remind you that NUMA and clusters are _two_ different things??? >> >>Or is it pointless??? >> >>Clusters have _zero_ shared memory. >> >> >>> >>>> >>>>> AFAIK their only plan is >>>>>to build a 50 tflop machine for 50 million dollar and the rest is unclear. >>>> >>>>Nothing is unclear at all. Oak Ridge is one of our national labs, just like Los >>>>Alamos, Lawrence-Livermore, etc. They always strive to "push the envelope" >>>>although this world of "clusters" is not as interesting as the old days of "real >>>>super-computers"... They have applications that run for months at a time. And >>>>they'd like to increase the size of the data but that would push the >>>>applications into years of run-time. Faster processing demands is what is >>>>driving this. >>>> >>>>But anyone could find that out by looking around or asking...
This page took 0 seconds to execute
Last modified: Thu, 15 Apr 21 08:11:13 -0700
Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.