Boinc Credits Question

Message boards : Number crunching : Boinc Credits Question
Message board moderation

To post messages, you must log in.

AuthorMessage
Profile MattDavis
Volunteer tester
Avatar

Send message
Joined: 11 Nov 99
Posts: 919
Credit: 934,161
RAC: 0
United States
Message 11744 - Posted: 24 Jul 2004, 4:16:04 UTC
Last modified: 24 Jul 2004, 4:16:34 UTC

Once again this isn't a Seti crunching question, but people here seem to be in the know :)

I've read two conflicting reports from people on threads about the following issue, and was wondering if anyone here *really* knew.

Some people say that different projects (Seti, Predictor, etc) use different credit standards so you can't compare/combine credits given on different projects. However, some people say they DO have similair standards so that you CAN combine/compare credits earned on various projects.

Anyone know?
ID: 11744 · Report as offensive
Profile AthlonRob
Volunteer developer
Avatar

Send message
Joined: 18 May 99
Posts: 378
Credit: 7,041
RAC: 0
United States
Message 11765 - Posted: 24 Jul 2004, 4:53:37 UTC - in response to Message 11744.  

> Anyone know?

They *could* have different standards, but right now they're (they being P@H and S@H) fairly normalized. Pirates and CPDN aren't really providing any credits to speak of yet.

Rob
ID: 11765 · Report as offensive
Profile Keck_Komputers
Volunteer tester
Avatar

Send message
Joined: 4 Jul 99
Posts: 1575
Credit: 4,152,111
RAC: 1
United States
Message 11851 - Posted: 24 Jul 2004, 8:45:02 UTC

That's one of the nice things about BOINC the credit system is supposed to be standardized so that the work done for different projects can be compared.

John Keck
BOINCing since 2002/12/08
ID: 11851 · Report as offensive
Profile MattDavis
Volunteer tester
Avatar

Send message
Joined: 11 Nov 99
Posts: 919
Credit: 934,161
RAC: 0
United States
Message 11920 - Posted: 24 Jul 2004, 13:25:00 UTC

I guess people are leaning towards a standard credit system.

I ask because I have my computers split between Seti@home and Predictor (and will split them again when another scientific, non-profit project becomes available like Einstein@home), and I want to be able to look at both credit totals and have a good feel of the proportion of work my computers do for either project. If Seti and Predictor had different credit standards, there would be no way to do this by looking at credit totals.
ID: 11920 · Report as offensive
Mark Stevens

Send message
Joined: 15 Feb 01
Posts: 28
Credit: 177,705
RAC: 0
United Kingdom
Message 11935 - Posted: 24 Jul 2004, 14:26:42 UTC

The problem with this credit system, as it stands at least, is that is doesn't seem to be that accurate unless I missed something fundamental here?

I have a variety of systems in my little seti farm ranging from a 600Mhz system to some twin 3GHz systems. The 600Mhz system crunches slightly under 2 WUs a day the twin 3GHz system over 16. However, after nearly 4 weeks the average credits for the slowest is 74 and 313 for the fastest.

I don't know if there's a good reason for it but it seems quite a way off compared to what I'd expect it to be.

Anyone else with results like these?

Mark


ID: 11935 · Report as offensive
CyberGoyle
Avatar

Send message
Joined: 2 Jun 99
Posts: 160
Credit: 3,622,756
RAC: 26
United States
Message 11939 - Posted: 24 Jul 2004, 14:39:12 UTC
Last modified: 24 Jul 2004, 14:39:36 UTC

There are several dynamics at work, the largest of which is that Berkeley scheduler has been doing all it can to keep up. I had a 647mhz PC outperforming a 3.2Ghz HT PC by a wide margin - until yesterday after Berkeley left the project offline so results could catch up.

There are other FAQ's on how credit is calculated, but it relates to your PC benchmarks as well as the results from two other people. The credit that is granted may not always be based directly on your own performance. Hunt the forums, the answer is there, I just forgot where it is.


<a>
ID: 11939 · Report as offensive
Mark Stevens

Send message
Joined: 15 Feb 01
Posts: 28
Credit: 177,705
RAC: 0
United Kingdom
Message 11962 - Posted: 24 Jul 2004, 16:29:11 UTC - in response to Message 11939.  

Thanks, I'm glad I'm not the only one with disproportionate results.

Yes, I've seen all the talk and description of WU and credits. Just doesn't seem to entirely fit the spec yet.

When I saw my slowest PCs also had benchmarks which outstrip the relative performance with my dual cpu systems I thought it was a little odd. eg. my 1Ghz laptop is 75% that of my 2x3Ghz system??? That's in fp and integer.

In real benchmarks, like Sandra, they come out with more realistic figures eg. more or less in ratio to the clock speed. My other solo cpu PC is similar. The twin systems just stick out as being somewhat odd in terms of benchmarks and credits.

Guess it's still early days and they've had their hands full with other more important issues...

ID: 11962 · Report as offensive
CyberGoyle
Avatar

Send message
Joined: 2 Jun 99
Posts: 160
Credit: 3,622,756
RAC: 26
United States
Message 12019 - Posted: 24 Jul 2004, 18:49:54 UTC - in response to Message 11962.  

> When I saw my slowest PCs also had benchmarks which outstrip the relative
> performance with my dual cpu systems I thought it was a little odd. eg. my
> 1Ghz laptop is 75% that of my 2x3Ghz system??? That's in fp and integer.
>


Regarding benchmarks, re-run the CPU benchmark on your dual CPU system. BOINC will return the results PER CPU, but your webpage PC summary will show the same numbers without mentioning it is per CPU.


<a>
ID: 12019 · Report as offensive
Mark Stevens

Send message
Joined: 15 Feb 01
Posts: 28
Credit: 177,705
RAC: 0
United Kingdom
Message 12033 - Posted: 24 Jul 2004, 19:18:19 UTC - in response to Message 12019.  

> Regarding benchmarks, re-run the CPU benchmark on your dual CPU system. BOINC
> will return the results PER CPU, but your webpage PC summary will show the
> same numbers without mentioning it is per CPU.

Yeah, been there too. My current benchmarks are the best of the crop and seem fairly typical for this sort of system. I know what you mean though. Although the fp seems fairly constant at around 2000, the integer test varies by over 100% on my system. 2000 to around 4600 I think the highest was? 800fp and 1400int are the 600MHz benchmarks btw.

Of course all are done without any other bg tasks active which is the norm for my systems anyway (just light surfing, editing and compiling).

ID: 12033 · Report as offensive
Profile MattDavis
Volunteer tester
Avatar

Send message
Joined: 11 Nov 99
Posts: 919
Credit: 934,161
RAC: 0
United States
Message 12037 - Posted: 24 Jul 2004, 19:23:50 UTC

Don't benchmarks automatically run once a week? I think I read that somewhere.
ID: 12037 · Report as offensive
Profile Thierry Van Driessche
Volunteer tester
Avatar

Send message
Joined: 20 Aug 02
Posts: 3083
Credit: 150,096
RAC: 0
Belgium
Message 12080 - Posted: 24 Jul 2004, 20:28:28 UTC - in response to Message 12037.  

> Don't benchmarks automatically run once a week? I think I read that
> somewhere.

Yes.
It is done automatically each 5 days.
ID: 12080 · Report as offensive
Darren
Volunteer tester
Avatar

Send message
Joined: 2 Jul 99
Posts: 259
Credit: 280,503
RAC: 0
United States
Message 12083 - Posted: 24 Jul 2004, 20:36:31 UTC - in response to Message 12080.  

> > Don't benchmarks automatically run once a week? I think I read that
> > somewhere.
>
> Yes.
> It is done automatically each 5 days.

However, this is only done when the program starts if it has been 5 or more days since they were last run (on the linux cli anyway). If you leave the software running, it does not interrupt to run the benchmarks after 5 days. If you never shut it down and restart it, you'll never run new benchmarks.

I would assume the windows version works this way also, but I don't know for sure.


ID: 12083 · Report as offensive
Profile Thierry Van Driessche
Volunteer tester
Avatar

Send message
Joined: 20 Aug 02
Posts: 3083
Credit: 150,096
RAC: 0
Belgium
Message 12090 - Posted: 24 Jul 2004, 20:51:10 UTC - in response to Message 12083.  

> However, this is only done when the program starts if it has been 5 or more
> days since they were last run (on the linux cli anyway). If you leave the
> software running, it does not interrupt to run the benchmarks after 5 days.
> If you never shut it down and restart it, you'll never run new benchmarks.

> I would assume the windows version works this way also, but I don't know for sure.

Yes, you're right. In Windows also, it runs only if the GUI has been shut down.

And I think Benchmark will run also automatically if a new software version of S@H has been downloaded.


ID: 12090 · Report as offensive

Message boards : Number crunching : Boinc Credits Question


 
©2024 University of California
 
SETI@home and Astropulse are funded by grants from the National Science Foundation, NASA, and donations from SETI@home volunteers. AstroPulse is funded in part by the NSF through grant AST-0307956.