Author: Robert Hyatt
Date: 08:31:05 07/08/02
Go up one level in this thread
On July 08, 2002 at 00:53:46, Terry McCracken wrote: >On July 07, 2002 at 21:54:10, Robert Hyatt wrote: > >>On July 07, 2002 at 10:47:32, Terry McCracken wrote: >> >>>On July 07, 2002 at 03:14:41, Slater Wold wrote: >>> >>>>On July 07, 2002 at 03:08:32, Terry McCracken wrote: >>>> >>>>>On July 07, 2002 at 02:49:49, Slater Wold wrote: >>>>> >>>>>>On July 07, 2002 at 02:46:08, Terry McCracken wrote: >>>>>> >>>>>>>On July 07, 2002 at 02:13:33, Slater Wold wrote: >>>>>>> >>>>>>>>How many nps does Diep get on 60 CPUs? >>>>>>>> >>>>>>>>I imagine it's not exactly "fine tuned", seeing as how testing time was short. >>>>>>>>Any thoughts about this? >>>>>>>> >>>>>>>>Grats on the first round win! >>>>>>>> >>>>>>>> >>>>>>>>Slate >>>>>>> >>>>>>>Hi Slate. I thought Diep was running on a supercomputer capable of 1 Teraflop, >>>>>>>running on 1024 processors? >>>>>>> >>>>>>>http://www.chessbase.com/events/events.asp?pid=141 >>>>>>> >>>>>>>http://www.sara.nl/hpc.www/teras/description/ >>>>>>> >>>>>>>Terry >>>>>> >>>>>>They only allowed him to use 60 CPUs. Time on this machine is $30 an hour per >>>>>>processor. >>>>>> >>>>>>For 60 CPUs it would be $1800 an hour. >>>>>> >>>>>>For 1024 CPUs it would be $30,720 an hour. >>>>>> >>>>>>I took this all from Chessbase's website. Sorry, I don't have the link handy. >>>>> >>>>>Ah...I see, thanks! It would kick serious ass on 1024 cpus!:o) >>>>> >>>>>Terry >>>> >>>>Oh, and you think it won't on 60 CPUs?! >>> >>>I didn't say that. Of course it would be strong. Let's say it would have been >>>interesting if he had use of all that processing power. >>> >>>Terry >> >> >>The problem is that almost _everyone_ is making gross assumptions about how >>easy it is to use 1024 processors. Or even 64. Take it from someone that >>has been doing this for _years_. It is not easy. Throw in NUMA and it >>becomes even harder... >> >>This is what used to really gripe me about comments made about Cray Blitz. >> >>"oh, you just had hardware faster than everybody else's..." >> >>Which was true, of course. But _using_ that hardware took years of effort to >>get things reasonably optimized... > >Sorry Robert! It wasn't meant to sound like this....sigh... > >Terry I wasn't really directing my response to you specifically. It was more of a "comment in general". Using the machine Vincent had was not going to be as easy as he thought. I tried to warn him...
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.