Author: Eugene Nalimov
Date: 19:36:18 03/01/00
Go up one level in this thread
On March 01, 2000 at 22:17:30, Robert Hyatt wrote: >On March 01, 2000 at 20:48:05, Pete Galati wrote: > >>On March 01, 2000 at 20:22:09, Robert Hyatt wrote: >> >>>On March 01, 2000 at 14:43:39, Pete Galati wrote: >>> >>>>On March 01, 2000 at 07:37:55, Graham Laight wrote: >>>> >>>>>On February 29, 2000 at 17:32:29, Robert Hyatt wrote: >>>>> >>>>>>On February 29, 2000 at 11:40:46, Ed Panek wrote: >>>>>> >>>>>>>On February 29, 2000 at 08:42:36, Robert Hyatt wrote: >>>>>>> >>>>>>>>On February 29, 2000 at 01:13:38, Georg Langrath wrote: >>>>>>>> >>>>>>>>>I tink that you can measure the speed of a analyze in nods per second. When will >>>>>>>>>a pc be comabarable with Deep Blue with that increasing in hardware every year >>>>>>>>>that is now? I think that it must be so some time in future. >>>>>>>>> >>>>>>>>>Georg >>>>>>>> >>>>>>>> >>>>>>>>Not easy to answer, but I would guess that the speed of deep blue is about >>>>>>>>1,000 times faster than the fastest program of today, based on the fastest >>>>>>>>program going 1M nodes per second, while DB could peak at 1B nodes per >>>>>>>>second. It averaged about 200M, but then it also had some complex eval stuff >>>>>>>>that would slow that 1M nps program down by a factor of 5-10 probably >>>>>>>> >>>>>>>>If you assume 1000x, with a doubling of machine speed every year (which is >>>>>>>>very optimistic) then it will take about 10 years to catch up. >>>>>>>> >>>>>>>>all of that analysis has lots of assumptions, however... >>>>>>> >>>>>>> >>>>>>> >>>>>>>Unless there is some incredible watershed breakthrough in processor technology >>>>>>> >>>>>>>Ed >>>>>> >>>>>> >>>>>>True. But I have been involved in computing since 1968, and there has been >>>>>>no "incredible watershed breakthrough in processor technology" for the past 32 >>>>>>years. Nothing suggests (to me) that one is forthcoming within the next 10+ >>>>>>years. >>>>> >>>>>There are companies out there making multi-processor machines in a low cost way. >>>>>What is required is not so much a technology breakthrough, but a marketing >>>>>breakthrough. Multi-processor computers needs to become both a big market and a >>>>>competitive market. >>>>> >>>>>Pentium processors are a big and competitive market. Trouble is, I don't think >>>>>they're the best architechture to put together in large numbers on the same >>>>>motherboard. >>>>> >>>>>Hey people - lets all find good reasons to need lots of processing power, stop >>>>>buying Pentiums, standardise on a multiprocessor archtechture, and start buying >>>>>it in large numbers! >>>>> >>>>>-g >>>> >>>>Ok, you got a few extra bucks on you that we can all borrow? Wouldn't I have a >>>>Quad Xeon if I could afford one? My 586 is old and slow because I don't have >>>>the money to replace it, truth is I'd be thrilled to have a 350mhz computer >>>>right now. So there is that money factor. >>>> >>>>But yeah, they don't put together large numbers of multi-processor machines >>>>because most people have no use for one, and that "most people" is what pays >>>>their bills. Us computer Chess fans are just another flicked bugger to computer >>>>manufacturers in general, but a good specialized market. >>>> >>>>Pete >>> >>> >>>Actually the number of dual-cpu machines is quite enormous. I have seen >>>some eye-popping numbers quoted by MB manufacturers... One day the quads >>>will get 'there'. >> >>I'd like to see that day. Any idea how many quad machines are in use by members >>here at CCC? >> >>Pete > > >I have 9 quad xeons at my office, plus the quad p6. :) Bruce has one. Amir >uses one. I just taught an undergraduate class in parallel programming, and >out of 15 students, three had dual-processor machines. You can put together >a good dual for 500-700 bucks. Near my office there is a large hall filled with 4- and 8-way SMP systems, and I regularly use one of them (usually to debug a program). Eugene
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.