Credits Given per WU

Message boards : Number crunching : Credits Given per WU
Message board moderation

To post messages, you must log in.

AuthorMessage
Luke
Volunteer developer
Avatar

Send message
Joined: 31 Dec 06
Posts: 2546
Credit: 817,560
RAC: 0
New Zealand
Message 781967 - Posted: 13 Jul 2008, 2:19:42 UTC

Hi Everyone
I do not seem to understand how credit is given per every WU... I have also done a bit of searching through my pending credit and have discovered these main types of WU's:
19 credits = seem to be shorter WU's running in at about 500-100 seconds on my machine.
50 credits = Running in @ 2500-3300 seconds.
52 credits = Appear to be the same as WU's with 50 given credits.
63 credits = Takes about 4000-4500 seconds.
78 credits = Takes the same time as WU's with 63 given credits...

Not all of this seems to match up by time, is there another factor involved that determines the outcome and granted credit?

Luke.

- Luke.
ID: 781967 · Report as offensive
W-K 666 Project Donor
Volunteer tester

Send message
Joined: 18 May 99
Posts: 19064
Credit: 40,757,560
RAC: 67
United Kingdom
Message 781975 - Posted: 13 Jul 2008, 2:43:52 UTC

The thread Estimates and Deadlines Revisited will maybe give you some insight into the non-lineararity of the AR/CR curve.

Credits are calculated from the number of Floating Point operations (FPops or FLops) performed to process the unit. But FLops are not equal, think difficulty Add and Multiply Versus Divide and Sqrt.
Also the processing performed at different AR's is different, e.g. Gaussians are only looked for in the central part of the AR curve.
ID: 781975 · Report as offensive
Luke
Volunteer developer
Avatar

Send message
Joined: 31 Dec 06
Posts: 2546
Credit: 817,560
RAC: 0
New Zealand
Message 782060 - Posted: 13 Jul 2008, 6:49:29 UTC - in response to Message 781975.  

The thread Estimates and Deadlines Revisited will maybe give you some insight into the non-lineararity of the AR/CR curve.

Credits are calculated from the number of Floating Point operations (FPops or FLops) performed to process the unit. But FLops are not equal, think difficulty Add and Multiply Versus Divide and Sqrt.
Also the processing performed at different AR's is different, e.g. Gaussians are only looked for in the central part of the AR curve.


Thanks for the thread WinterKnight....

Luke.
- Luke.
ID: 782060 · Report as offensive
W-K 666 Project Donor
Volunteer tester

Send message
Joined: 18 May 99
Posts: 19064
Credit: 40,757,560
RAC: 67
United Kingdom
Message 782085 - Posted: 13 Jul 2008, 8:41:56 UTC

Another link, to another of Joe's post, Variation in Requested Credit
ID: 782085 · Report as offensive
Profile Clyde C. Phillips, III

Send message
Joined: 2 Aug 00
Posts: 1851
Credit: 5,955,047
RAC: 0
United States
Message 782273 - Posted: 13 Jul 2008, 17:57:31 UTC

The object to the game is to get constant credits per hour, provided that the processor is giving full devotion to Seti (or other Boinc projects). Unfortunately that's impossible because it's harder to perform some types of flops than others. But, I see, at least for my machines, that credit-per-hour has been more constant the past couple of months, maybe since I had started using Alex Kan's cruncher. It looks like the high-credit workunits (like 75 to 100) and the low credit units (like 12 to 19) give the most credit-per-hour now. It used to be that the 25-to-30 credit units gave poor credit per hour. That's not true anymore. Results may be different with other (like AMD) processors.
ID: 782273 · Report as offensive

Message boards : Number crunching : Credits Given per WU


 
©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.