After a few days I now checked on the tasks and saw a significant increase of runtimes while the credits dropped almost in the same amount.
For example this task, almost 10 hours and 134,39 credits, that's 14,03 per hour!
Five days ago I had this here, around 5 hours and 103,64 credits, that's 20,90 per hour.
I'm now having tasks running at 53% with already 13 hours and up, and it seems they even get longer and longer.
Why are they getting so long suddenly?
I guess I'm outta here if the credits for them are so low, that damn CreditScrew is a PITA...
Hope you guys get the GPU-app running soon.

---

Life is Science, and Science rules. To the universe and beyond
Proud member of BOINC@Heidelberg
My BOINC-Stats

Hope you guys get the GPU-app running soon.

GPU apps are available for linux-x86_64+cuda, linux-x86_64+opencl_amd, and win_64+nvidia_opencl: https://numberfields.asu.edu/NumberFields/apps.php

Just enable "Run test applications" in your project preferences.

---

GPU apps are available for linux-x86_64+cuda, linux-x86_64+opencl_amd, and win_64+nvidia_opencl: https://numberfields.asu.edu/NumberFields/apps.php

Oh, looks like I missed that somehow then, my bad...
Sorry for opening the thread, maybe someone can delete it.

---

Life is Science, and Science rules. To the universe and beyond
Proud member of BOINC@Heidelberg
My BOINC-Stats

You can blame any credit issues on CreditNew. Credits seem to start higher and then drop off over time as the stats stabilize.

There are a mixure of both fast and slow WUs. The slower ones average about 2 hours on a newer 4GHz processor (slower with hyper-threading).

What about those who can crunch only on CPU? Can app be optimised further to decrease calculation time?
If not will it still be worth the hassle?

What about those who can crunch only on CPU? Can app be optimised further to decrease calculation time?
If not will it still be worth the hassle?

Further optimization of the CPU code might be possible, but it would require much work and chances are the performance gain would be minimal. The only library the algorithm relies on is gmp, and that is a very efficient library, so that leaves the discriminant algorithm as the only place where optimizations could be made. I am using the standard sub-resultant algorithm for the discriminant which I have further optimized for degree 10.

I'm not sure what you mean by "hassle". Although the GPU apps are faster (results vary from user to user, mine is ~5x faster); the majority of the results being returned right now are from CPUs. So CPU crunchers are very important to this project - if they all thought it was a hassle this project would cease to exist.

Thank you reasurring reply, Eric. I'm happy to hear that CPU crunchers are important to this project.

Hassle wasn't the right word to use. What I had on mind was that some users claim that crunching on CPU might become obsolete considering time needed to complete WU and credit given for valid task. So if I was going to process less tasks on CPU it would make me feel I'm not making any real contribution to the project.