credit/hour how much is possible??

Message boards : Number crunching : credit/hour how much is possible??

To post messages, you must log in.

Previous · 1 · 2 · 3 · 4

AuthorMessage
SekeRob

Send message
Joined: 7 Sep 06
Posts: 35
Credit: 19,984
RAC: 0
Message 28545 - Posted: 26 Sep 2006, 13:36:41 UTC - in response to Message 25514.  

how much you get in one hour, please report here
------------------------------------------------


I think the new system is pretty good....in taking last 10 (that's all i can see), i'm running on a stock benchmark of 8.10 per hour but am getting 9.15 on average.

Comparing WCG site (mix of UD / BOINC agents and any BOINC agent only running site i know, think it to be a big omission that no cumulative 'valid' tally is kept of units crunched as well as CPU time invested.....all i see is my total credit and RAC.

Coelum Non Animum Mutant, Qui Trans Mare Currunt
ID: 28545 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
zombie67 [MM]
Avatar

Send message
Joined: 11 Feb 06
Posts: 316
Credit: 6,347,141
RAC: 1
Message 28571 - Posted: 27 Sep 2006, 4:01:21 UTC - in response to Message 28545.  
Last modified: 27 Sep 2006, 4:03:09 UTC

Comparing WCG site (mix of UD / BOINC agents and any BOINC agent only running site i know, think it to be a big omission that no cumulative 'valid' tally is kept of units crunched as well as CPU time invested.....all i see is my total credit and RAC.


Look harder. =;^)

Click on "My Grid" (top)

Click on "Device Manager" (left)

Click on "Results Status" (left)

There ya go....

Edit: P.S. WCG takes the BOINC credits and multiplies by 7 to compete evenly with the UD clients. That is why your WCG stats are 7x higher than what shows on BOINCstats.
Reno, NV
Team: SETI.USA
ID: 28571 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Martin P.

Send message
Joined: 26 May 06
Posts: 38
Credit: 168,333
RAC: 0
Message 28969 - Posted: 6 Oct 2006, 12:49:53 UTC

On my G5/Dual 2.7 GHz I get 8.2 cr/hour and core. This is appr. half as much as comparable Intel CPUs gain.

ID: 28969 · Rating: -1 · rate: Rate + / Rate - Report as offensive    Reply Quote
Mats Petersson

Send message
Joined: 29 Sep 05
Posts: 225
Credit: 951,788
RAC: 0
Message 28970 - Posted: 6 Oct 2006, 13:23:55 UTC - in response to Message 28969.  

On my G5/Dual 2.7 GHz I get 8.2 cr/hour and core. This is appr. half as much as comparable Intel CPUs gain.


And half of comparable AMD CPU's.

It's been discussed before, and it's fairly clear that the PPC/Mac version of Rosetta isn't quite as performant as the x86 version. It's quite likely that this requires the compiler settings to be tweaked (or the compilers optimization stage). There's nothing obvious in the compiler settings for the PPC/Mac version tho', so it's not just a case of tweaking the makefile to compile with a different setting... Maybe if you can find someone to sponsor the GCC team to improve the PPC port of GCC - assuming of course it's not just simply a case of "PPC is bad at complicated floating point" - I know little about this processor architecture, so I can't really comment on that.

As "Darwin" OS stands for about 5% of all the OS's present in Rosetta (9910 out of around 195000) and 5% of the total credit produced, doubling the performance of this, assuming the number of users stay the same would add 5% to the overall work performed. Improving the Windows version of Rosetta by 0.2% would give roughly the same result - something that is probably a whole lot easier to achieve too.

--
Mats


ID: 28970 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
FluffyChicken
Avatar

Send message
Joined: 1 Nov 05
Posts: 1260
Credit: 369,635
RAC: 0
Message 28976 - Posted: 6 Oct 2006, 15:39:15 UTC - in response to Message 28970.  

On my G5/Dual 2.7 GHz I get 8.2 cr/hour and core. This is appr. half as much as comparable Intel CPUs gain.


And half of comparable AMD CPU's.

It's been discussed before, and it's fairly clear that the PPC/Mac version of Rosetta isn't quite as performant as the x86 version. It's quite likely that this requires the compiler settings to be tweaked (or the compilers optimization stage). There's nothing obvious in the compiler settings for the PPC/Mac version tho', so it's not just a case of tweaking the makefile to compile with a different setting... Maybe if you can find someone to sponsor the GCC team to improve the PPC port of GCC - assuming of course it's not just simply a case of "PPC is bad at complicated floating point" - I know little about this processor architecture, so I can't really comment on that.

As "Darwin" OS stands for about 5% of all the OS's present in Rosetta (9910 out of around 195000) and 5% of the total credit produced, doubling the performance of this, assuming the number of users stay the same would add 5% to the overall work performed. Improving the Windows version of Rosetta by 0.2% would give roughly the same result - something that is probably a whole lot easier to achieve too.

--
Mats



Though improving the PPC version (maybe just using a different compiler ? I don't know) may make it easier to get a good working XBOX360/PS3 since they are based around the PPC architecture iirc. We may also be able to get NASA to run it on the rovers ;-)
Team mauisun.org
ID: 28976 · Rating: -1 · rate: Rate + / Rate - Report as offensive    Reply Quote
Martin P.

Send message
Joined: 26 May 06
Posts: 38
Credit: 168,333
RAC: 0
Message 30503 - Posted: 2 Nov 2006, 13:04:42 UTC - in response to Message 28970.  
Last modified: 2 Nov 2006, 13:05:38 UTC


As "Darwin" OS stands for about 5% of all the OS's present in Rosetta (9910 out of around 195000) and 5% of the total credit produced, doubling the performance of this, assuming the number of users stay the same would add 5% to the overall work performed. Improving the Windows version of Rosetta by 0.2% would give roughly the same result - something that is probably a whole lot easier to achieve too.

--
Mats


No wonder that the PPC version is so slow. If all Rosetta programmers do the math like you do the results are random numbers anyway.

ID: 30503 · Rating: -1 · rate: Rate + / Rate - Report as offensive    Reply Quote
Mats Petersson

Send message
Joined: 29 Sep 05
Posts: 225
Credit: 951,788
RAC: 0
Message 30507 - Posted: 2 Nov 2006, 15:16:46 UTC - in response to Message 30503.  


As "Darwin" OS stands for about 5% of all the OS's present in Rosetta (9910 out of around 195000) and 5% of the total credit produced, doubling the performance of this, assuming the number of users stay the same would add 5% to the overall work performed. Improving the Windows version of Rosetta by 0.2% would give roughly the same result - something that is probably a whole lot easier to achieve too.

--
Mats


No wonder that the PPC version is so slow. If all Rosetta programmers do the math like you do the results are random numbers anyway.


Sorry....
1. Yes, I'm wrong, it should say "Adding ~6% improvement to the Windows platform should give the same result" (90% of the platforms are Windows, so a 6% improvement should give 5.4% more score, correct?). And I still think that's more easily achieved than doubling the performance of the PPC version - unless you know some direct improvements to GCC that you could supply that will automagically generate twice as fast floating point-code in GCC - but why hasn't that been implemented already, if that's the case. I don't have a PPC machine, and I haven't tried to analyze the PPC code for optimization potential (in fact, I've _NEVER_ used a PPC-based machine, or looked at optimizing code for it).

2. I'm NOT a Rosetta programmer per se. I've been looking at some of the code for performance optimisation potential. I may not be a math genius, but I do know how to optimise code to do the same calculation faster...

--
Mats
ID: 30507 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
zombie67 [MM]
Avatar

Send message
Joined: 11 Feb 06
Posts: 316
Credit: 6,347,141
RAC: 1
Message 30539 - Posted: 3 Nov 2006, 1:44:20 UTC

Depends on your goal.

If you want to squeeze the most performance out of your mac (G4, G5, or even intel), switch to SETI@home, using alexkan's optimized application. His application does not increase credit claims. It actually allows macs to crunch WUs faster. Just take a look at the top machine stats. Mac's own it. Used to be the Power Macs (G5), but the new Mac Pro's (intel) are quickly taking over.

But if your goal is to contribute as much to rosetta as possible, then run your G4 and be glad that even the small amount of results contribute to the whole, and are appreciated.

For me, I recently switched from having all machines crunch for all my projects, to having my macs running SETI, and the rest of the PCs running everything else. That way the tasks are assigned to the machines that do them best, and you get the most total work done (assuming SETI is part of your project portfolio). You get the most credits that way too. =;^)
Reno, NV
Team: SETI.USA
ID: 30539 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile Jack Shaftoe
Avatar

Send message
Joined: 30 Apr 06
Posts: 115
Credit: 1,307,916
RAC: 0
Message 30611 - Posted: 4 Nov 2006, 17:28:27 UTC - in response to Message 25514.  
Last modified: 4 Nov 2006, 17:34:39 UTC

E6600 @ 2.63ghz = 22.18 c/h/core

4400+ Toledo @ 2.42ghz = 14.25 c/h/core

Xeon 5160 @ 3.0ghz = 20.14 c/h/core


Relatively small sample of 8 workunits at 8 hours each.

Team Starfire World BOINC
ID: 30611 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Previous · 1 · 2 · 3 · 4

Message boards : Number crunching : credit/hour how much is possible??



©2022 University of Washington
https://www.bakerlab.org