My GTX460 just finished one of these wu's. Like always it took about 23 and a 1/2 hours. However, the amount of credit awarded is far less than the usual 67.500. Are the credits being recalculated for this type of wu or is it some kind of error? Thanks

http://www.gpugrid.net/result.php?resultid=4687075

Should have been 45k x 1.5 = 67.5 since received-sent is (barely) within 24h. Edit: a strange interference occurred with another results. The BOINC thing copied the lower credit of the other result. We'll try to see if a fix is possible for this case.

T

me too.

http://www.gpugrid.net/workunit.php?wuid=2940618

---

http://www.rechenkraft.net

Norman: yours is a different case - was returned in a short time window around 14/dec when we were fixing the credit calculations.

Should have been 45k x 1.5 = 67.5 since received-sent is (barely) within 24h. Edit: a strange interference occurred with another results. The BOINC thing copied the lower credit of the other result. We'll try to see if a fix is possible for this case.

T

Thanks Toni for your quick response.

OK.
thx Toni.

---

http://www.rechenkraft.net

Could someonem please explain to me the 24hr rule and why it justifies an increase in credit? I have received 45000, 56000 and 67500 for various completed "GIANNI" wu's.

The rule is that if you return the result within 24 hours you get a 50% bonus. Why? Well... why not?

The rule is that if you return the result within 24 hours you get a 50% bonus. Why? Well... why not?

Well there is a bit more to it than that :-)

The results from one set of wu are used to create the next set. That is why the project rewards fast returns. The sooner they can get enough returned they can use the results to create the next set.

You can exclude the "long" wu if you have a slower graphics card so that you have a better chance of getting them completed within 24 hours. Check your account preferences on the web site. Another thing to do is run with a minimum cache so you don't have wu sitting in "ready to start" status.

Does this 24 hr rule apply to all Long wu's or just the "GIANNI"wu's

All tasks of all types and size.

---

FAQ's

HOW TO:
- Opt out of Beta Tests
- Ask for Help

My concerns are that if credits are awarded based on an estimate of the amount of flops based on number of iterations, (as far as I am aware all BOINC projects calculate in a similar manner) and we enhance the credits based on a particular time frame to suit the project, which has nothing to do with that, are we "BOINC legal"?
I am not trying to put a fly in the ointment but it may appear to other projects that we are not playing by the rules.

At GPUGrid tasks are generated from the results of other tasks. Therefore the project needs a rapid turn around. You could just as equally consider the 24h time as 100% and take off 25% for a <2day return and 50% for a >2day but <5day return! Then there is a long task bonus, as this also helps the project.

This project rewards similarly to some other GPU projects. The maths based projects are the real culprits as their credit systems take no account of app complexity, and throw crazy credits GPU's.

As far as I am aware, few if any projects Fully implement the latest incarnation of the credit system, and it has not been calibrated. Unless all projects do it, the system is pointless.

Task credit should be awarded across every project in the same way (or as close as possible), despite the CPU type or GPU type; if a task takes 1h on a CPU and gets 50credits, it should get 50 credits on all devices (various CPU's, GPU's, multiple GPU's or combinations of CPU and GPU) no matter how long it takes to run the task.

If a project wants to average it's task credit to some extent, say by rewarding fast return and reducing credit for slow return, then I think there should be some discretion allowed; slow returns slow the research down, and fast returns expedite the research.

---

FAQ's

HOW TO:
- Opt out of Beta Tests
- Ask for Help

Many thanks
How do I reduce the cache to its minimum

Boinc Manager (Advanced view),

For Boinc 6.10.x,
Advanced, Preferences, and set the Additional work buffer to 0.1 or 0.01

For Boinc 6.12.x or Alpha versions,
Tools, Computing Preferences, and set the Additional work buffer to 0.1 or 0.01, but keep the connect every setting to above 0.1.

---

FAQ's

HOW TO:
- Opt out of Beta Tests
- Ask for Help

Thank you MarkJ and skgiven for being available at this holiday time it is a pleasure being part of arguablly one of the most worthwhile/best projects within BOINC

My GTX550TI can do the GIANNI WU in just under the 24 hours to qualify for the full bonus. :)

Boinc Manager (Advanced view),

For Boinc 6.10.x,
Advanced, Preferences, and set the Additional work buffer to 0.1 or 0.01

For Boinc 6.12.x or Alpha versions,
Tools, Computing Preferences, and set the Additional work buffer to 0.1 or 0.01, but keep the connect every setting to above 0.1.

I set the "connect every" to 0.1 and have zero in the "additional work buffer". I have the network activity always available. When a wu gets to about 95% done it requests a new wu, so it's got another one by the time the current wu has finished.

I upgraded my video card recently and saw my daily stats shoot through the roof. Then I realized I was getting the long WU's. I turned GPUgrid off and just ran seti for a day to see what kinda stats I would get. GPUgrid definately seem to give a significant bonus over other projects. I guess my only concern is the cobblestone was meant to be a reflection of work done... giving bonuses kinda skews things a bit. Maybe the who cobblestone thing is broken anyways. If you want such fast return times why not just set short deadlines? Do people care enough about these things that a credit war could start between projects who add bonuses to try and attract people? I just pick interesting projects. Give them similar priorty and let them go.

Granite -

In the past we tuned WUs so that they were in line (actually, lower than) with SETI @ home. (We still use the same credit factor for short WUs.) However, things became complicated over time as various things changed - most notably other projects were significantly over-granting credits. More recently, SETI completely overhauled the credit system (NewCredit). Most projects, including ours, did not adhere to NewCredit for a variety of reasons. Finally, last time I checked NewCredit's GPU factors still had to be tuned (in non-trivial ways).