NEW!!! ClimatePrediction.Net Weekly Stats !!!!

Toggle sidebar Toggle sidebar
BlackMountainCow

BlackMountainCow

Diamond Member
May 28, 2003
5,759
0
0
Hey all! This is a new stats report on a weekly basis for CLIMATEPREDICTION.NET



Group rank: 31 (- 1 last week: 30 ) :(

Members: 17 with 37 machines ( + 1 machine for idwxman) :)

Producing Members: 9 :(


Total Model Years: 1609.851 ( + 118.174 years) :Q

Team Ahead: Team Scotland ........... 1687.293 ... (+ 77.442 years)
Team Behind: Team Muropaketti ..... 1554.307 ... (- 55.544 years)


Full Runs: 18 ( + 2 ) :)

1 Full Run = 45 Model Years = :clock:
5 Full Runs = 225 Model Years = :beer:
10 Full Runs = 450 Model Years = :light:



Model Years + (Rank Gain or Loss):
  1. idlorj:...........................420.73 :beer::clock::clock::clock::clock:
  2. idwxman:.....................267.592 :beer:
  3. thehive:.......................240.059.........(+1) :beer:
  4. gistech1978:...............228.181..........(-1) :beer:
  5. teal97:.........................75.646 :clock:
  6. beatle:.........................66.27 :clock:
  7. Christian_Diepold:.......53.766 :clock:
  8. slangnrox:...................45.008 :clock:
  9. ta_andy:.....................41.9
  10. mwardscuf:.................40.015...........(+1)
  11. mardeth:.....................38.761...........(-1)
  12. darkwing:....................33.762
  13. smokeball:...................23.757
  14. vss1980:.....................16.881
  15. themagicman:.............15.024
  16. johnmal:.......................0 :(
  17. winstonsmith:...............0 :(

I know this is still very basic but as this is my first stat I will work on improving it week by week. Any comments or additions welcome!!
 
BlackMountainCow

BlackMountainCow

Diamond Member
May 28, 2003
5,759
0
0
@Overkiller: thx 2 ;)

I'm going to add some links an additional stuff in the next week's edition ... there r some very good stat sites out there!
 
BlackMountainCow

BlackMountainCow

Diamond Member
May 28, 2003
5,759
0
0
Originally posted by: Idlorj
Thanks for the stats.

I hope other people will join the team.
Click to expand...

Hehe! U as our CPN number ONE r surely glad for the stats!? ;)

But I really hope that it might attract more ppl to CPN, as we are losing ground fast to other teams and our producing members are getting fewer and fewer as well. I mean, we're just 17 and of those 17 only 9 produced last week!! :(

 
C

Coquito

Diamond Member
Nov 30, 2003
8,559
1
0
just wondering why it seems that climate prediction isn't listed in the welcome thread at the top of the forums where all other dc projects are.

i'm interested in the project:)
 
O

Overkiller

Platinum Member
Feb 22, 2003
2,461
0
0
BlackMountainCow..what kind of cpu power are we talking about...is the project at such a state that 1 medium grade pc can make a huge impact? honest answers please :D ;)
 
ericlp

ericlp

Diamond Member
Dec 24, 2000
6,137
225
106
I downloaded the program but can't really get it to work... What I got from the User Forums @ climateprediction... Oh well, Nice try I guess but not really worth it if one can't really get the perfromance out of a Chip like Seti can. As to the "mods" reply ... all I can say is Lazy!



Hyperthreading was invented to raise the efficiency of the CPU internal parts (let other Threads use the unused units) because one inefficient and crappy binary (this programm is an excellent example) alone doesn´t make much use of them (even the extra 256KB Cache of a Barton doesnt raise the Programs efficiency).
The Code is that crappy that there must be speedups from Hyperthreading alone by min 50%, with real a Dualprocessor probably 80% or more.
It is easy to split the world into a north and south part and let 2 threads compute them.
It has an efficiency of about 20% while other realworld distributed projects reach 80%.
This is the P4 Version of the Program to Monitor the CPU efficiency:
ftp://ftp.heise.de/pub/ct/ctsi/perfwtch.zip
The Athlon version was sent by mail to me after i asked the Author.

This project does an enormous amount of calculations and takes huge amounts of time/electricity and they don´t even use Compiler optimizations, i simply can´t understand it.
Punch their nose into the problem and they won´t even look.

If a program that long running isn´t worth all possible optimizations (min base P6 optimization and multithreading which will probably give over 300% speedup together) then it isn´t worth using my CPU time.

I probably let it run until 90% phase 3 to look what the world will look like and then wipe it from the Harddisk forever.

Back to top


Deamiter
Moderator



Joined: 22 May 2003
Posts: 203
Location: Minnesota, USA
Posted: Tue Jul 01, 2003 5:39 am Post subject:

--------------------------------------------------------------------------------

You have to understand that in many DC projects, hyperthreading is next to useless. As each calculation is small and directly dependant on the one before it, it often takes more time to split up the calculation than to just send it through all in one piece. While there ARE some chip optimizations that are not being used, I'd give the client a POSSIBLE 25% speedup, but only after a year or so of working on optimization. Not the 300% that would be true if the program was just hacked out in C++ by some 1st year college kids.

There are always ways to make the program faster, but our discussion of the optimization is suggestive rather than necessary. In many cases, implimenting a certain feature can take longer than the time that is saved (working on the science to make the overall data more efficient will save much more time in the long run than any 0.3% increase in speed.)

 
BlackMountainCow

BlackMountainCow

Diamond Member
May 28, 2003
5,759
0
0
@Overkiller:

Well, that IS a good question. :D Honestly, you should have a good machine to run it on. For example: I run it on an Athlon XP +2000 and it takes me about 45 days to complete a full run = 45 model years. Though, I got to say that this is my main machine, so per day there r about 4 hours when the machine is occupied with games, compiling or zipping ... So IF u run it 24/7 without disturbance on an XP +2000 I guess u end up with round 42-40 days for one full run. Honest enough? ;)
 
BlackMountainCow

BlackMountainCow

Diamond Member
May 28, 2003
5,759
0
0
Originally posted by: Coquito
just wondering why it seems that climate prediction isn't listed in the welcome thread at the top of the forums where all other dc projects are.

i'm interested in the project:)
Click to expand...

NO idea why it is not up there. Maybe because it has been running for three months only so far. But in those 3 months over 882272 Model Years were crunched!! So that's not bad I guess ;)

Nice to hear, that u r interested in it. It's really a good thing I guess, though it takes much more time per WU than SETI or any other DC project I've known so far.
 
BlackMountainCow

BlackMountainCow

Diamond Member
May 28, 2003
5,759
0
0
Originally posted by: ericlp
I downloaded the program but can't really get it to work... What I got from the User Forums @ climateprediction... Oh well, Nice try I guess but not really worth it if one can't really get the perfromance out of a Chip like Seti can. As to the "mods" reply ... all I can say is Lazy!

Hyperthreading was invented to raise the efficiency of the CPU internal parts (let other Threads use the unused units) because one inefficient and crappy binary (this programm is an excellent example) alone doesn´t make much use of them (even the extra 256KB Cache of a Barton doesnt raise the Programs efficiency).
The Code is that crappy that there must be speedups from Hyperthreading alone by min 50%, with real a Dualprocessor probably 80% or more.
Click to expand...

Well, u have to keep in mind that CPN has gone public just 3 months ago. So I don't think u can expect it to be running as smooth and perfect as other DC projects that have been around for several years now.

It is easy to split the world into a north and south part and let 2 threads compute them.
Click to expand...

I doubt that u can split a world into North and South on a climate model! I mean wind and weather just don't stop because they step over the equator. So I guess that s splitting would make no sense at all IMHO.

It has an efficiency of about 20% while other realworld distributed projects reach 80%.
Click to expand...

Where do u get these numbers from?? I mean, i don't have a P4 so the HT is useless for me anyway but still, I wanna know where u got that numbers from?? 20%, I just don't believe that.

This project does an enormous amount of calculations and takes huge amounts of time/electricity and they don´t even use Compiler optimizations, i simply can´t understand it. Punch their nose into the problem and they won´t even look.
Click to expand...

Well, either u r a coder of the project or else how would u know? Did u take a look at the sourcecode? I doubt that, cause it's not public. So please bring some evidence for your statements. Also, did u ever ask those ppl or give them feedback? What I get from the forums they DO care about the project and take feedback very seriously. The DO looks at it.

I probably let it run until 90% phase 3 to look what the world will look like and then wipe it from the Harddisk forever.
Click to expand...

Well, do what u wanna do ...


I just can't believe that u have such a negative opinion of a DC project that is hardly out of the cradle. And just because ONE of their moderator made a somewhat fuzzy answer u condemn the whole project ... For example i had many probs with F@H at first cause the server to upload my results was down so i had severl WUs in queue that almost expired. But do i condemn F@H for that ... NO, i don't.


My 2 cents ...
 
I

Idlorj

Member
Sep 17, 2003
32
0
0
From some of the discussion on the Climate Prediction Boards the primary program is over 500,000 lines of Fortran code. This program was not originally designed for use by PCs.
 
ericlp

ericlp

Diamond Member
Dec 24, 2000
6,137
225
106


I just can't believe that u have such a negative opinion of a DC project that is hardly out of the cradle. And just because ONE of their moderator made a somewhat fuzzy answer u condemn the whole project ... For example i had many probs with F@H at first cause the server to upload my results was down so i had severl WUs in queue that almost expired. But do i condemn F@H for that ... NO, i don't.


My 2 cents ...[/quote]

Hmmm, That was not me talking I just posted it from what I could find on the Climate Forums... Since I guess it doesn't do HT, then I won't be running it. As I agree with the poster. 2.4 Ghz With out HT on SETI will Process about 8-9 WU a day but with HT On it will do 13-14UW a Day. I use two HT CPU's to do just that so that is easly 20% faster. OK, it's only 3 months out... YEAH??? Hello? Intel is not making it difficult to code for HT. I am not a coder I am a user and I guess if I got a few machines with HT then why would I want to GO SLOWER with the eqipment I got? Why not wait a few more months and optimize the code for people with faster chips? Must be the UK way...
 
BlackMountainCow

BlackMountainCow

Diamond Member
May 28, 2003
5,759
0
0
@ericlip

Well, I don't think this is gonna lead anywhere near a solution. All I can say is that I WILL stick to CPDN and continue to post my weekly stats. I respect your opinion though I don't share it at all. I don't know why u don't like CPDN, but hey, so be it.

I'm just wondering if u'll run it when or if they ever implement HT compatibility. I agree that the coders should use any optimization possible. But that would mean an extra client for Athlon64s as well. And for HT P4s. And for other x86s like Athlon XP and P4 w/o HT.

As far as I can see they released a stable version that runs well on every platform. As time goes by they will increase their coding capabilities and the optimizations. That's no doubt for me. If they didn't they'd vanish from the DC scene as quickly as they appeared. But that won't happen ...
 
BlackMountainCow

BlackMountainCow

Diamond Member
May 28, 2003
5,759
0
0
Joining TA is just as easy as with other projects. When registering u just pick TA as team.

HT is not supported so far as to use both virtual CPUs for one project. But u can run 2 instances of it at once. Hopefully there's a real HT support to come!
 
You must log in or register to reply here.

TRENDING THREADS

COMPANY
RESOURCES
FOLLOW
Top Bottom