FYI, Some preliminary results showing (Granted) Credit vs. CPU time for three modestly-sized crunching systems. Your experiences? Thanks!

FYI, Some preliminary results showing (Granted) Credit vs. CPU time for three modestly-sized crunching systems. Your experiences? Thanks!

If the data weren't splattered all over the place, someone might be able to draw a conclusion from that. As it stands, it looks like a demonstration of chaos theory.

---

Proudly Banned from Predictator@Home and now Cosmology@home as well. Added SETI to the list today. Temporary ban only - so need to work harder :)



"You can't fix stupid" (Ron White)

FYI, Some preliminary results showing (Granted) Credit vs. CPU time for three modestly-sized crunching systems. Your experiences? Thanks!

If the data weren't splattered all over the place, someone might be able to draw a conclusion from that. As it stands, it looks like a demonstration of chaos theory.

You might check out the new & improved graph in this message.

I do not crunch much SETI/SETI Beta (1 WU/week or less) so I might be wrong - but it seems to me that SETI will increase their credits with 5.17 (current Beta client), the fixed credits of the 5.17 stock client are very close to those that CC 4.19 claims for the results (by benchmark) and thus quite high compared to 5.15.

SIMAP reduced their credits with HMMER - it's very low at the moment but they are still in the process of finding a way to measure the work done.

It looks as if it will still take quite some time until projects are comparable.

I would like to say a big thank you to everyone who contributed to this thread. It is clear that a lot of effort has gone into making the system fairer, and a lot of effort here on Rosetta estimating the practical impact of the changes.

Clearly not everyone agrees with exactly how to make the transition to the new system, but what is impressive is the amount of unity over how the credits get calculated in future.

River~~

---

OK folks, it's been sometime since my last update. Many projects have been added and more results have been added to the projects I've previously collected data on. My P4 1.8 has died and beeen ressurected as the AMD X2 4800 you now see. All data is from standard Boinc clients. This time I've included the data from optimized applications to the charts/tables below. The individual xls files and charts are available for download/inspection upon request. There is one xls file for each project and one worksheet for each host.

I'll draw no conclusions here and simply present what I'm seeing, leaving it up to you to fish out what you may see. (I.E which are giving more/less than what the benchmark would indicate, and which are credit rich/richer)

This data is constantly changing and being updated. I wanted to wait till I was done, but quite frankly the more I do, the more I see needs to be done (like breaking project app versions down), and I'm beginning to think it'll never end.

OK folks, it's been sometime since my last update. Many projects have been added and more results have been added to the projects I've previously collected data on. My P4 1.8 has died and beeen ressurected as the AMD X2 4800 you now see. All data is from standard Boinc clients. This time I've included the data from optimized applications to the charts/tables below. The individual xls files and charts are available for download/inspection upon request. There is one xls file for each project and one worksheet for each host.

I'll draw no conclusions here and simply present what I'm seeing, leaving it up to you to fish out what you may see. (I.E which are giving more/less than what the benchmark would indicate, and which are credit rich/richer)

This data is constantly changing and being updated. I wanted to wait till I was done, but quite frankly the more I do, the more I see needs to be done (like breaking project app versions down), and I'm beginning to think it'll never end.

But what is your conclusion ?

---

Team mauisun.org

OK folks, it's been sometime since my last update. Many projects have been added and more results have been added to the projects I've previously collected data on. My P4 1.8 has died and beeen ressurected as the AMD X2 4800 you now see. All data is from standard Boinc clients. This time I've included the data from optimized applications to the charts/tables below. The individual xls files and charts are available for download/inspection upon request. There is one xls file for each project and one worksheet for each host.

I'll draw no conclusions here and simply present what I'm seeing, leaving it up to you to fish out what you may see. (I.E which are giving more/less than what the benchmark would indicate, and which are credit rich/richer)

This data is constantly changing and being updated. I wanted to wait till I was done, but quite frankly the more I do, the more I see needs to be done (like breaking project app versions down), and I'm beginning to think it'll never end.

But what is your conclusion ?

The conclusion is.....something which comes at the end. An end I don't see happening unless I just stop. LOL

There are many things that can be gotten from this. It all depends on what you endevour to research. To me the more I do the more I see needs being done, more data collected. More specific data as well (i.e division by wu types etc). I've posted this on many websites for both management to get an idea about what their individual project looks like in "the grand scheme of things", and for participants also.

To me, I think most projects are doing a fine job, but there's always room for improvement, refinement, etc. Each project has it's own peculiarities which they can deal with. Einstein, for example, grants more credit/hour on thier short wus vs. the long ones and there's diversification within those two groups. Einstein is clearly the "king of credit" when viewed by those running standard clients. The Boinc Simap 5.08 hmmr for winxp is clearly the lowest, followed closely by VTUathome, and then Seti (enhanced 5.15). Of those projects with Fixed Flops count and/or running averages Rosetta is the closest to achieving parity if the benchmark is the benchmark. Seti is low, Einstein is high.
The projects that still use a benchmark based credit system and a quorum are still being skewed by the use of third party boinc clients. The severity of the skewing depends on the saturation of third party boinc clients within those projects and the frequency that those will be matched up and the higher claims granted. (One could check that saturation rate, by looking at a decent number of results and checking the WU ID of each participant in the quorum. I haven't the time ATM)
One thing that my charts won't show is a loss of parity by projects who don't use a quorum and grant whatever is claimed. My charts would show that project at perfect parity (or nearly so) since I'm getting what I claim.

I think each project has it's own issues to deal with. I think Boinc/projects need to decide just what should be the value of credit and what should be "the norm". This is not easy to do given all the variables.