Your opinions please...

Here's mine: I think we're still getting too much credit for the cpu time spent.

Regards,

Sysfried

---

Happy member of Team: Planet 3D Now!

I'm getting quite a bit more than claimed on my Core Duo running Vista. My top claim was 49, but most top out in the 30-32 range.

Same host running Linux is similar. Looks like it's getting mostly "short" results. It's claiming around 16 for those.

My P4 (which should be taken with a large grain of salt due to heat issues) isn't even claiming 60. So it's also getting more granted than claimed.

---

Kathryn :o)
The BOINC FAQ Service
The Unofficial BOINC Wiki
The Trac System
More BOINC information than you can shake a stick of RAM at.

I guess a lot depends on the project you want to compare it to. I do a lot of what I call Credit Test Wu's for the other Projects, I just do 1 Wu to compare the Credits against this Project. I can actually get more Credit @ some of the other Fixed Credit Projects Per Hour of work, I also get less at some of the other same type Projects.

It also depends on what you compare. The base line version (typically Windows 32-bit), or the version that gives highest credits (here Linux 64-bit). For most 32-bit Windows hosts the credit here will indeed be much higher than on most other projects. For the 64-bit client it's not exceptional.

---

BOINC.BE: For Belgians who love the smell of glowing red cpu's in the morning
Tutta55's Lair

Wasn't the credit here set a bit higher to account for WUs that failed to award credit for whatever reasons?

---

Dublin, CA
Team SETI.USA

Wasn't the credit here set a bit higher to account for WUs that failed to award credit for whatever reasons?

Yup, that's been said - besides, as you can read on announcements, there will be new very short WUs with no credit at all... so in the end, credit awarding should be similar to other projects ;)

---

La Tierra de un Dios que no supo aceptar / su falso derecho a la libertad - Tierra Santa
Descárgate mi primer eBook Amaneceres

Why not make it so it has to be a quorom of 3 and then take the claimed credit that is between the high and the low. I am taking a look at some of my results right now and there is a quorom of 2 and my claimed credit is about 32 and my granted is 60. I know this is a set at 60 for each one basically. But in the future maybe think about making it a quorom of 3 and taking the middle claimed as the total granted to all three. I hope this makes sense.

Ni

Please no! That reduces our crunching capacity by 50%. If the science integrity is sufficient with a quorum of two, then leave it there.

---

Dublin, CA
Team SETI.USA

Please no! That reduces our crunching capacity by 50%. If the science integrity is sufficient with a quorum of two, then leave it there.

Agreed with that. Also, the sooner benchmark based crediting becomes a thing of the past in Boinc projects, the better IMO.

---

BOINC.BE: For Belgians who love the smell of glowing red cpu's in the morning
Tutta55's Lair

Please no! That reduces our crunching capacity by 50%. If the science integrity is sufficient with a quorum of two, then leave it there.

Agreed with that. Also, the sooner benchmark based crediting becomes a thing of the past in Boinc projects, the better IMO.

Hah, but there is that difference between benchmark credits (client), credits by fpops (application) and credits set by the server, you know? ;-)

I like them set by the server, then those benchmarks can go to hell... well, not really, they're still useful to see if someone's computer isn't underclocking. :)

---

This maybe a stupid newbie question but here goes.

Why is every WU I submit getting 60 credits?
No matter what the claimed credits are everything gets 60 credits.

This maybe a stupid newbie question but here goes.

Why is every WU I submit getting 60 credits?
No matter what the claimed credits are everything gets 60 credits.

It's because that's what the Server set's the Credit at ... :)

Right now it is set at 60 because the program does not correctly identify the total amount of time it has taken to finish a WU. If something interrupted the 'thinking' of the computer, the time used by the pc might be reset. That is why you will notice a wide range of times to complete a WU.

---

Here's a sample of my recently crunched WUs in some projects. Cosmo gives the most per hour, followed by SIMAP and Einstein. I usually claim about 8.3, my puter is a AMD AthlonXP2200+, running Suse Linux 10.1.

Project      Time       Claimed    Granted    C/h cl.   C/h gr.
Cosmology: 143:29:45    1171,69    2340,00     8,17     16,31
SIMAP:      39:19:42     326,65     600,63     8,31     15,27
Einstein:  108:07:03    1503,28    1503,28    13,90     13,90
Docking:   110:22:26    1038,09    1336,50     9,41     12,11
ABC:        35:11:22     288,76     422,16     8,21     12,00
CPDN:     2388:55:14   23846,40   23846,40     9,98      9,98
Rosetta:   213:15:05    1737,58    1858,59     8,15      8,72
RCN:        83:36:28     682,13     682,13     8,16      8,16
Malaria:   157:32:38    1281,39    1272,28     8,13      8,08

I could easily do with a lot less crdits, even 35 would be nothing to complain about for me.

---

Grüße vom Sänger

As far as I can see the only project more generous with the credits than C@H is QMC. But I prefer the science of C@H and anyway going for the most generous project around feels like cheating. So I'll stick with the second most generous. ;-)

It is strange how things come and go though. Not that long ago Einstein was one of the most generous of projects. Then it became one of the meanest to the extent that I stopped running it for the sake of my team's standing as well as my own. Now, from Saenger's data, which I'm sure is correct, it's relatively generous. Rather than all this trial & error and constant moving of goal posts, you'd think that there would be some generally agreed objective way of measuring the amount of work involved in processing a work-unit?

---

Rather than all this trial & error and constant moving of goal posts, you'd think that there would be some generally agreed objective way of measuring the amount of work involved in processing a work-unit?

Mega- Credit is that elusive dark matter that you can never quite get a handle on in Boinc. My guess is it will always be this way unless 1 formulae is used on ALL projects.

---

Rather than all this trial & error and constant moving of goal posts, you'd think that there would be some generally agreed objective way of measuring the amount of work involved in processing a work-unit?

Mega- Credit is that elusive dark matter that you can never quite get a handle on in Boinc. My guess is it will always be this way unless 1 formulae is used on ALL projects.

But even the definition of "fair" can be defined in different ways.
The old benchmark method took the potential power of the puters and the time, they were used as a measure, regardless of the amount of work done.
The new serverside credits takes actual work done for the project as the measure.

A certain algorithm and it's implementation in machine code may be of different efficiency on various machines, depending on the OS/CPU architecture. With the benchmarks the various runtimes and credit claims for the same WU will usually lead to undergranting for the setups that suit better, as the lower value will be granted. With the serverside credits better suited setups will get more credits/h, and the project has to make a decision which setup they will take as the standard, so either some will get less then usual or some will get more.

Here I obviously get more then usual, so I presume my OS/CPU-setup suites this project extraordinary fine and some other setup is used as the gold standard.

Both definitions of "fair" are good and valid, I don't think one could be ruled out on fairness reasons, but serverside credits are less manipulatable, so I would prefer them. And if a project will be suited better for Win/Intel I will get less then usual and can decide wheather to crunch for the project in spite of this because I think it's worth it, or if I look for some better suited one.

---

Grüße vom Sänger

Saenger, your list compares the 32-bit Linux clients. It would be interesting to have a similar comparison for the 64-bit clients, which are becoming more pupular. For the Cosmology project we found that there was little difference between the 32 and 64 bit versions. For other projects, like ABC and RieselSieve there is a larger difference. Also SIMAP seems to have a pretty fast 64-bit version.

---

BOINC.BE: For Belgians who love the smell of glowing red cpu's in the morning
Tutta55's Lair

Saenger, your list compares the 32-bit Linux clients. It would be interesting to have a similar comparison for the 64-bit clients, which are becoming more pupular. For the Cosmology project we found that there was little difference between the 32 and 64 bit versions. For other projects, like ABC and RieselSieve there is a larger difference. Also SIMAP seems to have a pretty fast 64-bit version.

I will see that soon, as I've ordered a new puter. The mainboard has some difficulties to get here, but as soon as I have some new data I will post them.

---

Grüße vom Sänger

Saenger, your list compares the 32-bit Linux clients. It would be interesting to have a similar comparison for the 64-bit clients, which are becoming more pupular. For the Cosmology project we found that there was little difference between the 32 and 64 bit versions. For other projects, like ABC and RieselSieve there is a larger difference. Also SIMAP seems to have a pretty fast 64-bit version.

I will see that soon, as I've ordered a new puter. The mainboard has some difficulties to get here, but as soon as I have some new data I will post them.

First seven are through on the new machine (C2D E6750 running kubuntu7.04), here are the credits (I've put the two still not granted ones on 50 granted as well, I think they will validate):

Time       Claimed    Granted    C/h cl.   C/h gr.
11:15:08    188,90     350,00     16,79     31,10

So I'm still well off with the 50 credits per WU, still get nearly double of what I ask for.

---

Grüße vom Sänger