![]() |
|
|
Thread Tools | Display Modes |
|
#1
|
|||
|
|||
![]() |
#2
|
|||
|
|||
![]() "Ray Vingnutte" wrote in message . .. http://news.bbc.co.uk/1/hi/technology/3983131.stm " ...yet the chip technology powering the machine is the kind which can be found in familiar devices like games consoles." It doesn't say how many they wired up in parallel, but building more powerful machines by wiring up more Pentiums in parallel is easy - the hard part is developing the software to make them appear one. The real development will be software structures which allow you to dd computing power by adding standard boxes to a LAN incrementally, in which case the limit to the "computer" speed will be the number of Dells you want to buy .... |
#3
|
|||
|
|||
![]()
On Sun, 7 Nov 2004 00:48:10 +1100
"Peter Webb" wrote: "Ray Vingnutte" wrote in message . .. http://news.bbc.co.uk/1/hi/technology/3983131.stm " ...yet the chip technology powering the machine is the kind which can be found in familiar devices like games consoles." It doesn't say how many they wired up in parallel, but building more powerful machines by wiring up more Pentiums in parallel is easy - the hard part is developing the software to make them appear one. The real development will be software structures which allow you to dd computing power by adding standard boxes to a LAN incrementally, in which case the limit to the "computer" speed will be the number of Dells you want to buy I remember reading some time ago that NASA used to(and still may I don't know) use what is called a Beowulf system, these I believe are basically off the shelf computers that are networked together. Whether that is still how things are don again I'm not sure. This might help explain better than I can http://www.beowulf.org/ ... |
#4
|
|||
|
|||
![]() Heres a little more info, seems to be a lot if you search google, I'll have to put that to one side for later looking at ;-) http://www.xyroth-enterprises.co.uk/bclust.htm ... |
#5
|
|||
|
|||
![]()
Peter Webb wrote:
"Ray Vingnutte" wrote in message . .. http://news.bbc.co.uk/1/hi/technology/3983131.stm " ...yet the chip technology powering the machine is the kind which can be found in familiar devices like games consoles." It doesn't say how many they wired up in parallel, but building more powerful machines by wiring up more Pentiums in parallel is easy - the hard part is developing the software to make them appear one. The real development will be software structures which allow you to dd computing power by adding standard boxes to a LAN incrementally, in which case the limit to the "computer" speed will be the number of Dells you want to buy ... Huh, that's an interesting way to architech a network. I wonder what the ROI is for something like this in a standard office where most users are only concerned with documents, spreadsheets and email apps. I think I'll take a look at this. |
#6
|
|||
|
|||
![]() "Peter Webb" wrote in message u... snip It doesn't say how many they wired up in parallel, but building more powerful machines by wiring up more Pentiums in parallel is easy - the hard part is developing the software to make them appear one. The real development will be software structures which allow you to dd computing power by adding standard boxes to a LAN incrementally, in which case the limit to the "computer" speed will be the number of Dells you want to buy snip IMHO, this will be the next revolution in computing. 64bit is just a temporary excitemet. BV. |
#7
|
|||
|
|||
![]()
Benign Vanilla wrote:
"Peter Webb" wrote in message u... snip It doesn't say how many they wired up in parallel, but building more powerful machines by wiring up more Pentiums in parallel is easy - the hard part is developing the software to make them appear one. The real development will be software structures which allow you to dd computing power by adding standard boxes to a LAN incrementally, in which case the limit to the "computer" speed will be the number of Dells you want to buy snip IMHO, this will be the next revolution in computing. Not really. Concepts like this are around and have been implemented for the last 20 years. The problem with that is the question "how to program those things". Another problem is: A great deal of real world problems aren't of the sort that parallel computing makes sense. The computer will spend more time in sending data packets to processing units then doing anything else. Or if the problem is really sequentially (meaning B can be computed only after A has been computed. But then C can only be computed after B's result has shown up etc), then no parallel computer in the world can come up with the result faster the a single CPU can. -- Karl Heinz Buchegger |
#8
|
|||
|
|||
![]()
no parallel computer in the world can come up with the result faster
the a single CPU can. what's a CPU? _______ Blog, or dog? Who knows. But if you see my lost pup, please ping me! A HREF="http://journals.aol.com/virginiaz/DreamingofLeonardo"http://journal s.aol.com/virginiaz/DreamingofLeonardo/A |
#9
|
|||
|
|||
![]()
"Twittering One" ha scritto nel messaggio
... no parallel computer in the world can come up with the result faster the a single CPU can. what's a CPU? Celestial Power Unleashed... Luigi Caselli |
#10
|
|||
|
|||
![]() "Karl Heinz Buchegger" wrote in message ... snip IMHO, this will be the next revolution in computing. Not really. Concepts like this are around and have been implemented for the last 20 years. The problem with that is the question "how to program those things". Another problem is: A great deal of real world problems aren't of the sort that parallel computing makes sense. The computer will spend more time in sending data packets to processing units then doing anything else. Or if the problem is really sequentially (meaning B can be computed only after A has been computed. But then C can only be computed after B's result has shown up etc), then no parallel computer in the world can come up with the result faster the a single CPU can. I guess I should have been more clear. My point being that someday, IMHO, Windows or Lindows, or Linux or BallmerOS, or whatever is dominant, will have a feature where you can share x% of your processor time, so as companies buy computers they will just be adding to their grid. Then in the off hours you can sell your bandwidth. This will all be part of the OS, and the need to handle it will be disconnected from the application. The OS will just use the available CPU bandwidth. BV. |
Thread Tools | |
Display Modes | |
|
|
![]() |
||||
Thread | Thread Starter | Forum | Replies | Last Post |
Who needs a time machine "information travels faster than time" | timothy liverance | Policy | 3 | May 18th 04 11:02 PM |
Who needs a time machine "information travels faster than time" | timothy liverance | Space Shuttle | 0 | May 18th 04 09:33 PM |
Who needs a time machine "information travels faster than time" | timothy liverance | Space Station | 0 | May 18th 04 09:32 PM |
At sea-level NOBODY has gone faster than 600KPH, but you believe that they reach 1200KPH, don't you? | bart janssens | Satellites | 84 | March 13th 04 08:28 AM |
Incorrect assumptions about the speed of light | Arobinson319 | Amateur Astronomy | 16 | September 29th 03 05:04 PM |