A reset from 60 to 50 per WU seemed fair enough. BUT it seems to have coincided with the average WU taking much longer than before. So for my Q6600 the points per hour per cpu have dropped from about 35 (excellent) to 20 (mediocre). Anybody else noticed this?

---

A reset from 60 to 50 per WU seemed fair enough. BUT it seems to have coincided with the average WU taking much longer than before. So for my Q6600 the points per hour per cpu have dropped from about 35 (excellent) to 20 (mediocre). Anybody else noticed this?

I thought it might just be me but I have also noticed what seems to be an increase in the time it takes to crunch a WU. Did the length change?

---

A reset from 60 to 50 per WU seemed fair enough. BUT it seems to have coincided with the average WU taking much longer than before. So for my Q6600 the points per hour per cpu have dropped from about 35 (excellent) to 20 (mediocre). Anybody else noticed this?

At first glance it has stayed the same here. Most WU still take between 6000 and 9000 seconds on my machine, a momentarily non OC'ed Q6600 equivalent.

---

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

I can't say that I've seen an increase in the Wu Lengths either, on my Quad Cores the Wu's take between about 40 min's up too a little over 2 hr's to do. I'm still seeing those sort of times as of now, it could be though that more of the 2+ hour Wu's are being sent or resent and that could make it seem like the Wu's have increased in length ...

WU parameters are generated (more or less) randomly. While run-times may fluctuate now and then, the average running time should remain nearly constant.

---

Scott Kruger
Project Administrator, Cosmology@Home

First seven are through on the new machine (C2D E6750 running kubuntu7.04),
[snip]

I've got some more by now, so on a wider database this is the result:
Wu-count: 88 (excluding the faulty ones)
run time: 114:06:48h, 410807,96sec
claimed credits: 2433,16 (21,32c(h)
granted credits: 4400 (38,56c/h)

If I just look at the last 20 it's even "better": 20,67claim, 45,84 grant.

Is this still a Linux anomaly or does this project grant too much throughout all OS?

My current ratios on this machine for the projects I ran in the last 4 weeks is this (in descending order):
Seti (opt): 48c/h, 5h runtime
QMC: 46c/h, 50h
Cosmo: 39c/h, 114h
CPDN: 35c/h, 195h
Einstein: 30c/h, 133h
SIMAP: 25c/h, 128h
ABC: 25c/h, 1,5h
malaria: 21c/h, 85h
WCG: 21c/h, 85h
XTreme: 21c/h, 2h
RCN: 20c/h, 13h
Lattice: 20c/h, 32h
Leiden: 15c/h, 2h
yoyo: 15c/h, 118h

Imho Simap to Lattice are OK, Leiden and yoyo too little (yoyo adjusted his credits, I have to crunch one of those still), QMC to Einstein are too much, Seti is optimised, they should incorporate that and adjust the credits as well.

---

Grüße vom Sänger

Scott should get it over & just cut the Credits to 10 per Wu, that way nobody could complain about getting to much Credit.

Saenger ...what I find interesting if I am reading your posts right is you were getting 31cr/hr back when Scott had it at 60cr/wu and now you are getting 39 cr/hr when the wu is only worth 50cr/wu.
How the heck did that happen? Must be that new fangled math. hehe

Imho parity between all projects are never going to happen kinda like saying their will always be rich,middle class,and poor people you will never see everyone middle class.

---

Saenger ...what I find interesting if I am reading your posts right is you were getting 31cr/hr back when Scott had it at 60cr/wu and now you are getting 39 cr/hr when the wu is only worth 50cr/wu.
How the heck did that happen? Must be that new fangled math. hehe

Imho parity between all projects are never going to happen kinda like saying their will always be rich,middle class,and poor people you will never see everyone middle class.

That was my old machine (AthlonXP23200+), now sitting with my parents, mostly off and not crunching any more. Initially the c/h-factor came down on that machine of course, my new machine (C2D E6750@3,6) is a lot faster, so a bigger c/h is to be expected and absolutely fine. The new machine is about two times as fast per core as my old one was on it's single one, so in the whole 4 times as fast.

A bit of imparity is fine with me, and if it's impossible to get the several OS/CPU setups in line, calibrate with the majority (fine for my Linux atm, hopefully not for longer and Littleweak will vanish). An imparity of +/-10% is fine, an imparity of more then 100% is to avoid.

I know that in the past Littleweak puters didn't run fine here, Linux were a lot faster, so they would get a lot more per hour. It's been stated, that the new code is better on the Gates-system, but I can't say, I don't use it. If my puter was the norm, it shoould probably be something between 30 and 25 per WU, but that's a bit too small sample imho ;)

---

Grüße vom Sänger

Remember, credits are intentionally on the high side, to account for WUs that failed to award credit for various reasons.

---

Dublin, CA
Team SETI.USA

Once the project is more stable (i.e. I don't have to cancel large batches of WUs without granting credit), I will probably lower the granted credit again. This won't happen for a while, though.

---

Scott Kruger
Project Administrator, Cosmology@Home

Once the project is more stable (i.e. I don't have to cancel large batches of WUs without granting credit), I will probably lower the granted credit again. This won't happen for a while, though.

Thanks Scott. I think it's probably not at the top of the priority list, especially considering things aren't near to settled yet. I had a 27k+ secs WU on a 3.0 Ghz P4 today, and that's *just how things are* in Alpha mode.

http://www.cosmologyathome.org/result.php?resultid=1035211

-jim

---

Once the project is more stable (i.e. I don't have to cancel large batches of WUs without granting credit), I will probably lower the granted credit again. This won't happen for a while, though.

I don't believe that the credit is to high on this project. I only average 23 cr/h on a Opteron 285 running Linux, which is nothing spectacular. QMC gives me far more. It is about the same as Einstein (22) and I consider it good reward for the work done.
When you start getting 9 to 16 as I do for LHC, Rosetta, Lattice you debate if it is worthwhile. Even 17 to 18 for CPDN on this machine is not huge.

With work units running from 4,000 to 12,200 seconds and 50 granted credits I get from 44 cr/h tp 14.5 cr/h. The 44 credit work units don't happen very often so that is why I get an recent average of 23 (averaged over 138 work units).

The longer the work unit the less I get.

People getting very high cr/h usually have very fast/recent machines.

Reducing the amount of credit is not the answer but awarding credits for the amount of work done (similar to a fixed system) would be a much better system I believe (the longer the work unit the more credit it earns).

Credit is the root of much turmoil in BOINC. I've been with BOINC for 2 years and every forum of every project I have joined has a discussion(s) of whether credit rewards are sufficient or fair. We all know that "fair" is a matter of opinion and there is apparently no satisfactory answer. Some want only credits and say as much, others "claim" to be only interested in the science, while others presumably like the best of both worlds. I'm the latter. I want to participate foremost in projects whose goals and objectives interest me. If a single credit was never forthcoming I would not lose any sleep over it. I'm human, so if credits are offered as part of the project package then I only ask that the credit scheme treat everyone as fairly as technologically and humanly possible giving the disparity of hosts and operating systems.
Further, who is to say one project is more important than another except the participant, so to take credit out of the equation it would be desirable for all projects to offer comparable credits, then people can really focus on participating in those projects of personal interest rather than who offers the most for the least.

I don't mind the current credit and I also wouldn't mind if you lowered it either. Either way I will be crunching. It would make more of a difference if the level of responsiveness changed. I have been extremely happy with this project because of the level of interaction that Scott and Ben have provided. That is worth more than the credits in my opinion. (credits don't hurt either ;) )

---

Nothing is set in stone, so I'll consider all opinions on this subject.

---

Scott Kruger
Project Administrator, Cosmology@Home

Scott..

After you have your application debugged, ported to all the major operating systems, optimized and stabilized, then you should take a good look at finding a way to determine credits for the WUs accurately and make the credits close to some standard, if you have the time and manpower. If you never find the time and manpower or the nature of this project's computations just don't allow for accurate estimates then just do it the quick and dirty way like you seem to be doing now. Or maybe throw together a script that would tell you how many credits some arbitrary average machine is earning rather than rely on such reports from crunchers (small sample)

This is your project, you call the shots, we're just here to help. The only things we have a right to ask for is to not have our precious resources wasted and communications from the project admins. You're doing just fine on those 2 points. Ignore the extremists who think you should make the credits your top priority. I think you'll find most of us think the credits should be your lowest priority.

Scott..

After you have your application debugged, ported to all the major operating systems, optimized and stabilized, then you should take a good look at finding a way to determine credits for the WUs accurately and make the credits close to some standard, if you have the time and manpower. If you never find the time and manpower or the nature of this project's computations just don't allow for accurate estimates then just do it the quick and dirty way like you seem to be doing now. Or maybe throw together a script that would tell you how many credits some arbitrary average machine is earning rather than rely on such reports from crunchers (small sample)

This is your project, you call the shots, we're just here to help. The only things we have a right to ask for is to not have our precious resources wasted and communications from the project admins. You're doing just fine on those 2 points. Ignore the extremists who think you should make the credits your top priority. I think you'll find most of us think the credits should be your lowest priority.

Hi Dagorath

This issue has been discussed long ago and it is Scotts' intent to award equal credit for equal work and it will be possible...there are currently different length workunits and will continue....it could be fixed credit or server assigned but from past discussions will probably not be benchmark based.

I hope this answers your questions.

---

Thanks for your timely and informative reply, JRenkar.

Lets be honest with ourselves. We are in it for the credit (I know I am.) My CPU are not particularly fast (none over or underclocked). I suppose it makes up for all the other projects where my computers only get a fraction of the credit to what they have claimed. I am sure that if the credit system changes to drastically, you will see many users stop crunching.