Post your views here, I'll do as well, but this is just a starter.

I think something went wrong with URL you posted, because i couldn't "come" here using link which you posted...

anyway, speaking of claimed credit, i have no objectives that some projects are giving relative more points than others, unless user know how it is granted and differences between projects aren't too big.

All links are broken by some software of Willy, probably to prevent spamming or whatever. You have to delete the blancs, wich show as %20 in my browser (german FF, Suse, ISO8859)

But on topic:

I don't have any objections about a credit deviation of about +/- 10%, but the use of the so-called "optimized" clients, that do nothing but fiddle with the benchmarks, is something completely different. They do nothing for the science, they only claim more credit than usual.

The admin of RCN stated it quite correct in his forum, and that's why I restarted my crunching over there.

If you have compiled an optimized science application, and use the "optimized" client for the compensation of the loss of credits/h through the original BOINC method of claiming, it's imho OK, but that's the exeption to the rule.

As BOINC is open source, and I want it to stay this way, perhaps it really is the best to put the credit stuff in the science app (if closed source) or entirely on the server. Unfortunately this is not always possible, so the quorum is another good method of restricting the abuse of "optimized" clients. As you can see, it's even posiible in projects with a quorum of 1, if the WUs are comparable enough, like it's done @Rosetta since today.

I think there are serious issues with the new Rosetta credit system. If you think optimised clients were bad, you ain't seen nothing yet!

Unless the project is prepared to scrutinise every claimed credit (and given the arguments raised by certain sectors of the Rosetta community about the incorrect reporting of cpu's, the "benefits" of tweaking ram and clocks etc etc, and the subsequent effects that this has on benchmarks - ignoring the fact that they are running optimised clients) I don't see how they can separate fact from fallacy.

Bottom line from a cynic, forget about cross project parity and be prepared for massive credit inflation.

As I see, Willy is her and reads it. I have a technical question:

If a project grants consistently credits, that are by a factor of X too high compared to the others, would it be possible for you to get them even here on the pages by dividing all numbers from the XML by X?

Possible but not wanted. It would take a cooperation between the major stats sites to do this here. If I would do this and the others don't then it will rain complaints here, people will say BOINCstats-statistics are incorrect, and leave to the other stats-sites (which is bad you know )

It's my opinion that it is up to the projects to regulate the granting of credits.

From the Rosetta forums - official credit announcement thread


Someone broke the VTU validator a short time back with "optimised" claims - not suggesting that this has happened at Rosetta, but it could happen.

Willy shouldn't be the arbitrator of BOINC projects.

We're the users, the volunteers, it's up to us.

Perhaps Willy could add a "hot air" factor to the projects statistic, if it really is possible to compute the rate of granted inflated credits (=" hot air" not science). :-)
Norbert

XML stats are just that they don't report who uses what BOINC version

That's right, but there was a discussion over @Rosetta to grant all users the inflated credits according to the "optimised" clients. I think it's a bit over now, as no retroactive adjustment of the stats to work done can be done any longer, so a discussion of the direction of this adjustment is obsolete as well That's mine as well, but if only one project would behave different, and it's too nice/important/valuable/whatever to ditch it totally from the stats, it would be a possibility.

A little late to the party here, but just saw SIMAP was moving to fixed credit.

Got a question, but first.. If anyone here has extensive time in Folding@Home, I think issues that people raise here against fixed credits gets totally disproven there. Some cores are better optimized for different CPU's, ie, Tinker's are better on AMDs, QMD's on Intel. All machines get a mix of WU's, and over the long run, points per given unit of time is dependent entirely on the raw power of your system. Never once in the couple years I folded my way to 100k at F@H for the top team [H]ardOCP was there a problem over granting credit. And that put to rest the debate of "its for the science, not the competition"; F@H would have much, much lower production if not for the insanely fierce competition between the top two teams especially, but also all teams. They overclocking to the brink (responsibly, stability checking), and they "borg" every computer they can get their hands on. Just like in the economy, competition, enabled by fair credits, is GREAT for science! If all that is good enough in the F@H experience, why is there even so much debate in the BOINC-o-sphere?

Anyway. My 02. I was wondering though, is there a list anywhere of known credit granting systems? I was thinking of voting with my CPU time and only working on projects with fixed systems. I know R@H is now along with CPDN. Any others?

Einstein is 1.

Anders n

Thanks, missed that one!

Got these two long-winded CPDN models winding up (about 2wks), then I'll be splitting 50/50 Rosetta & Einstein, or in thirds if there's another project awarding points in the same way. Thankfully, those are my two favorite projects anyway

Folding would be another one, if they decided to go further on BOINC, but they stopped at the beginning of this year to further develop the BOINC version. The last post on the forum there was that it will be only after the release of the PS3 and the GPU client.

Woah! It seems that the crediting has brought up many queries!!!

Quite interesting, just as a thought that the fixed credit system for wu's sounds the best in theory. A suggestion would be all project wu's have a deadline and if completed by then a fixed amount is awarded with maybe a little extra credit*time under the deadline maybe???

As for projects running on different make CPU's I find that my Pentium runs through ufliud wu's quite fast where as my old AMD or Laptop's AMD seem to struggle with them more thans it does on other projects (the thinking being my laptop could cope with the small ufluids wu's)

Yet a little puzzle is according to the credit comparision table it seems to indicate ufluids as a good creditor yet my graphs in the BOINC client show it rewarding roughly a 1/4 of what other projects are crediting?

Understandable there would be slight differences between each project but I've now put a hold on new ufliud wu's in favour of crunching for other projects as I try rack up a decent amount of credits. (Although I plan on crunching again for ufluids once my old AMD is back up as just a cruncher machine)

It is only a theory not aplicable to existing project...or not based on knowledge of project needs.

There a project with unpredictable running times - RCN from math area, BURP, Renderfarm and actually any rendering/reytracing/radiosity/any_similar to name just some. For some project there is no such a thing as same type of WU hence no fixed credit.
Also, I don't think you would like to wait a year before you get any credit from CPDN - when model is completed and validated; it is good to mention that deadline is no serious issue for CPDN. There are partial results (trickles) accepted and begin credited on-the-fly.

It is known that each CPU is doing a bit different on each project.
It is known that each OS - be it Win, Linux or MacOS is doing a bit different on each project.

It is known that each application version, possible optimalization - do to extra-level optimalization in source code, compilation, libraries etc. - brings a bit different speed. Now, we come to 32-bit and 64-bit version. Might be same hardware but different speed and effectiveness.

There is no simple answer like "grand fixed credit for same type of WU".

A valid point about the wu's, especially the likes of CPDN, I definately agree!

I guess it's back to the old saying "Don't fix it if it's not broken!!!"