Any idea why the score for seti@home on both intel and nvidia gpu's is ridiculously low? I mean crunching f.e. collatz conjecture gets me 200 to 500 times the amount of boinc credits for the same amount of crunching time on the same hardware. This makes the seti project veeery unattractive and I love this project as it was the first one I started (even from the time there was no boinc yet).

EDIT: even worse: the score for GPU is about the same as for CPU. Seems odd.

---

basically in the case of multibeam applications, the credit is normalised to the cobblestone scale, but the reference application is a CPU AVX enabled application. Buried underneath is the use of Boinc Whetstone, which is not a vectorised form. As most CPU's these days will have SSE or higher, and the applications optimised, the last 100 results for the CPU application that set these scales will 'claim' between ~1/2x to ~1/4x what they should be, then the GPU variants are normalised to that.

For astropulse there isn't an AVX stock version AFAIK, so that'll be roughly ~1/2x (50%)

Also there seems to be a lot of noise in the credit system, to the tune of > +/-30%. If you're like me, at odds with Murphy, you'll tend to receive on the lower end of that noise.

---

"Living by the wisdom of computer science doesn't sound so bad after all. And unlike most advice, it's backed up by proofs." -- Algorithms to live by: The computer science of human decisions.

Added to which, comparisons to other projects often fail because some projects have stopped even paying lip-service to the cobblestone scale. If the normalisation and variance issues Jason mentions are ever rectified, it might be possible to persuade other projects to return to the official definition of credit, so that comparisons once again become meaningful. But that time may still be some way off.

OK I can understand that there can be some deviation and up to 75% less credit is no problem for any truthful volunteer, but a factor of >99% is simply too great. This hurts Seti@home as some users will abandon the project simply for better boincstat scores.

OK I can understand that there can be some deviation and up to 75% less credit is no problem for any truthful volunteer, but a factor of >99% is simply too great. This hurts Seti@home as some users will abandon the project simply for better boincstat scores.

Well those that are after useless made up point may do that, but many users of SETI@home are here for the science to find ETI.

---

SETI@home classic workunits: 93,865 CPU time: 863,447 hours
Join the [url=http://tinyurl.com/8y46zvu]BP6/VP6 User Group[

Credit scores per task per task really cannot be compared between projects. Put it simply, some projects throw credits over the wall like confetti at a wedding, others give a fixed score per task, and others are even more parsimonious with credits than SETI. Use them to compare your performance over time within a project by all means, but don't expect to be able to compare between projects.

---

Bob Smith
Member of Seti PIPPS (Pluto is a Planet Protest Society)
Somewhere in the (un)known Universe?

It's not just between projects, credit screw is also extremly inconsistent within Seti.

While comparing my AP tasks with those from other users with a GTX 750 I discovered something really strange. This are results from one of those users:

http://i.imgur.com/7VYTz3Y.png

The first strange thing is that there are tasks with lesser runtime that give more credit then those with higher runtime. But I guess we've all seen that. But the second strange thing are the overall high scores. For comparison, here's an actual screenshot from my results:

http://i.imgur.com/O3kzkmr.png

I guess you can see the difference.

Now I'm in here for the science and I know those credits aren't worth anything, but still I'm curious. Any thoughts on this?

---