mersenneforum.org  

Go Back   mersenneforum.org > Great Internet Mersenne Prime Search > Hardware

Reply
 
Thread Tools
Old 2006-11-23, 08:01   #1
TehPenguin
 
TehPenguin's Avatar
 
Jun 2006
Perth, Western Australia

2×5 Posts
Default Intel's 80 core beast...

Hopefully the multithreading for Prime (http://mersenneforum.org/showthread.php?t=5313, see the comment about Version 26 by 'Prime95') will be working in next few years, cause I want one of these:

http://www.intel.com/pressroom/archi...0926corp_b.htm

(I wonder what complications may arise from running Prime on a Core 80?)
TehPenguin is offline   Reply With Quote
Old 2006-11-23, 13:10   #2
alpertron
 
alpertron's Avatar
 
Aug 2002
Buenos Aires, Argentina

22·3·113 Posts
Default

Quote:
Originally Posted by TehPenguin View Post
(I wonder what complications may arise from running Prime on a Core 80?)
This core 80 will not be compatible with the Intel Architecture Set, so the program will need to be rewritten from scratch in order to run on that chip. This means that there will not be any Prime 95 for the core 80.
alpertron is offline   Reply With Quote
Old 2006-11-23, 15:21   #3
Aillas
 
Aillas's Avatar
 
Oct 2002
France

2×5×13 Posts
Default

What about 80 instances of Prime 95, each running on a core
Aillas is offline   Reply With Quote
Old 2006-11-23, 21:01   #4
moo
 
moo's Avatar
 
Jul 2004
Nowhere

14518 Posts
Default

Quote:
Originally Posted by alpertron View Post
This core 80 will not be compatible with the Intel Architecture Set, so the program will need to be rewritten from scratch in order to run on that chip. This means that there will not be any Prime 95 for the core 80.
Read above
moo is offline   Reply With Quote
Old 2006-11-25, 02:15   #5
ixfd64
Bemusing Prompter
 
ixfd64's Avatar
 
"Danny"
Dec 2002
California

45168 Posts
Default

How long will it take to completely rewrite Prime95?
ixfd64 is offline   Reply With Quote
Old 2006-11-25, 07:41   #6
Uncwilly
6809 > 6502
 
Uncwilly's Avatar
 
"""""""""""""""""""
Aug 2003
101×103 Posts

2×4,787 Posts
Default

Quote:
Originally Posted by ixfd64 View Post
How long will it take to completely rewrite Prime95?
It would be better to run Mlucas or Glucas on this.
Uncwilly is offline   Reply With Quote
Old 2006-11-26, 09:55   #7
TehPenguin
 
TehPenguin's Avatar
 
Jun 2006
Perth, Western Australia

A16 Posts
Default

Quote:
Originally Posted by alpertron View Post
This core 80 will not be compatible with the Intel Architecture Set, so the program will need to be rewritten from scratch in order to run on that chip. This means that there will not be any Prime 95 for the core 80.
Excuse my naivety, but does this mean that all the OS's have to be adapted to use these CPU's as well?
TehPenguin is offline   Reply With Quote
Old 2006-11-26, 22:10   #8
TheJudger
 
TheJudger's Avatar
 
"Oliver"
Mar 2005
Germany

21278 Posts
Default

Quote:
Originally Posted by TehPenguin View Post
Excuse my naivety, but does this mean that all the OS's have to be adapted to use these CPU's as well?
Yes... but probhably you'll never see this CPU. Its just for testing interconnects between the cores... the cores itself are likely very simple (and slow compared to current CPUs).
TheJudger is offline   Reply With Quote
Old 2006-11-27, 02:26   #9
jasong
 
jasong's Avatar
 
"Jason Goatcher"
Mar 2005

3·7·167 Posts
Default

Quote:
Originally Posted by TheJudger View Post
Yes... but probhably you'll never see this CPU. Its just for testing interconnects between the cores... the cores itself are likely very simple (and slow compared to current CPUs).
I've wondered for a while whether or not this whole "add more cores to get more performance" could be bad news for primality determining programs. Is it not possible that in the cpu manufacturers quest to get more performance by adding cores that performance of individual cores could get worse instead of better? We've been hovering around 2 to 3 Gigahertz for a while now, through a few "generations" of hardware. Maybe as things progress, individual cores will have less performance individually, but more performance when you consider the entire chip.
jasong is offline   Reply With Quote
Old 2006-11-27, 07:49   #10
jinydu
 
jinydu's Avatar
 
Dec 2003
Hopefully Near M48

2·3·293 Posts
Default

Mr Woltman did say in another thread that he is working on a new version of Prime95 that will allow multiple processors to work on a single test:

http://www.mersenneforum.org/showthr...2432#post92432
jinydu is offline   Reply With Quote
Old 2006-11-27, 11:13   #11
TheJudger
 
TheJudger's Avatar
 
"Oliver"
Mar 2005
Germany

11·101 Posts
Default

Quote:
Originally Posted by jasong View Post
I've wondered for a while whether or not this whole "add more cores to get more performance" could be bad news for primality determining programs. Is it not possible that in the cpu manufacturers quest to get more performance by adding cores that performance of individual cores could get worse instead of better? We've been hovering around 2 to 3 Gigahertz for a while now, through a few "generations" of hardware. Maybe as things progress, individual cores will have less performance individually, but more performance when you consider the entire chip.
The per-core-performance increased. Compare an old Pentium 4 2.66GHz/533MHz FSB/512kB L2 Cache with an Core 2 Duo E6700 2.66GHz/1066MHz FSB/4MB shared L2 Cache. You'll see that the performance increased alot.

On the other hand the chip makers just "don't know what to do with these transistors".
The first Pentium 4 (Willamette, 180nm, 256k L2) has 42M transistors
The 2nd gen P4 (Northwood, 130nm, 512k L2) has 55M transistors
The 3rd gen P4 (Prescott, 90nm, 1M L2) has 125M transistors
The 3.5th gen P4 (Prescott 2M, 90nm, 2M L2) has XXX transistors
The 4th gen P4 (Cedar Mill, 65nm 2M L2) has XXX transistors

I'm not sure about transistor count of the last 2...
From Willamette to Northwood it was not much more than a simple shrink plus double L2 cache. You can say, that 256k L2 costs 13M transistors.
Willamette core logic (without cache): 29M transistors
Northwood core logic (without cache): 29M transistors
Prescott core logic (without cache): 73M transistors
(these numbers may not be very accurate but should be in the right magnitude)
In real world apps you'll see that the Prescott has an even worse IPC (Instructions Per Clock) than the Northwood while offering ~400MHz more clockrate.

When you build an chip you'll ask first: How many transistors can we use?
This was a limit for several years... but now they don't know what to do with them ==> hey, lets build dualcores/multicores ;)
TheJudger is offline   Reply With Quote
Reply

Thread Tools


Similar Threads
Thread Thread Starter Forum Replies Last Post
48-core chip - Intel hj47 Hardware 6 2009-12-20 13:59
Intel Quad Core optimization joblack Hardware 1 2009-06-11 02:42
Optimising work for Intel Core 2 Duo or Quad Core S485122 Software 0 2007-05-13 09:15
CGPU multi core x86, Intel Larrabee dsouza123 Hardware 2 2007-02-14 18:00
Intel Core Duo processor drew Hardware 5 2006-05-29 07:00

All times are UTC. The time now is 08:28.

Fri May 14 08:28:06 UTC 2021 up 36 days, 3:08, 0 users, load averages: 2.74, 2.59, 2.17

Powered by vBulletin® Version 3.8.11
Copyright ©2000 - 2021, Jelsoft Enterprises Ltd.

This forum has received and complied with 0 (zero) government requests for information.

Permission is granted to copy, distribute and/or modify this document under the terms of the GNU Free Documentation License, Version 1.2 or any later version published by the Free Software Foundation.
A copy of the license is included in the FAQ.