I have calculated my average granted credits over 145 send wu’s and I find it very low.
For those 145 wu’s I have claimed 6403,84 credits and I have only received 3690,52 granted credits , this is just a average off 57,6 %. Is this common ?. I find it quite low.

---

Hi

This is true for the most useres with fast machines.
I get from 50 WUs with claimed at 30 credits maybe 10 with my claimed credit or higher.
The most i get is around 20 credits because the best and the lowest claimed will be deleted.

You can read more about in Boinc Wiki.

greetz Mike

---

With each crime and every kindness we birth our future.

IMHO it's because of the low benchmarks for P4 HT machines... very low. They crunch WU quite fast, have low benchmarks so the claimed credit for those machines are LOW....

---

BOINC VALIDATOR and Credits

---

And the beat goes on
Sonny and Cher

BOINC Wiki

I have calculated my average granted credits over 145 send wu’s and I find it very low.
For those 145 wu’s I have claimed 6403,84 credits and I have only received 3690,52 granted credits , this is just a average off 57,6 %. Is this common ?. I find it quite low.

IMHO, 25.28 CREDITS/WU seems about right to me. I figure the average wu is worth about 25 credits. The credits you claimed of 44.16 CREDITS/WU seems a bit high to me. Or maybe I am expecting too little :D

If anyone has an opinion on how many credits an average WU is worth, I'd be glad to hear it.

-Mark

I have the feeling that over the last month or three there has been a fall in the "average" granted credit.
When I first started Seti Boinc, I seemed to get an average of about 30.
According to the "pending credit monitor" I now have an average of just over 20, on both machines.
Perhaps due in part to all the optimised clients on machines where the benchmark results in them very low credit, and pulling everyone else down?

However, some machines seem to claim very little credit.

For example this pc is 2.6GHz, 1GHz faster than mine, yet he claimed one third of the credit than I did for this wu. his benchmarks are almost the same as mine, if not slightly worse, yet the machine is clearly far faster. The result, we all got very little credit. The problem lies I suspect with the benchmarks, which are well known to be highly inaccurate.

Seeing as the benchmark only runs once every 5 days, why not make it a bit longer, so as to improve its accuracy? It takes about 30 seconds. Hardly long enough IMHO, to accurately judge a machines ability. Extending it would penalise no one, as we all run it.

As an aside I take just over 3 hours a WU, with a benchmark that suggests 5 hours. I claim 30, and get 20, but if I took as long as the benchmark suggest I would claim 50, and get 20. I.E. 40% of what I claim.
Screwy.
But all of us are equally effected, so it really makes not one bit of difference. We all get, I would imagine, the same average of about 20c/wu.

But I do think a more accurate benchmarking method is needed.

---

Foamy is "Lord and Master".
(Oh, + some Classic WUs too.)

Using benchmarks provides a verry inaccutare time to complete. Why they don't
use your average completion time is beyond me. Using a guesstimate when the have actual data to give out work & calculate a guessed at cred is just plain stupid.

As for credit, faster machines with optomised clients do get a bonus.
I'm using one and can crunch 3 times as many WU. Each claims an avg of 6 credits
but is granted 3-4 times that due to the "calculation method"

Bottom line I receive more credits for the same amount of computer time.

I have the feeling that over the last month or three there has been a fall in the "average" granted credit.
When I first started Seti Boinc, I seemed to get an average of about 30.
According to the "pending credit monitor" I now have an average of just over 20, on both machines.
Perhaps due in part to all the optimised clients on machines where the benchmark results in them very low credit, and pulling everyone else down?

The optimized clients are not the only reason. Even with "normal" clients faster computers do not claim as much credit as they should claim. And as the computers are getting faster and faster the average claimed credit gets lower and lower. And as the average claimed credit is decreasing, the average granted credit is also decreasing.

But this also means: It's not true that you get same credit for the same work. As time goes by you get less credit for the same work...

Carsten

I have calculated my average granted credits over 145 send wu’s and I find it very low.
For those 145 wu’s I have claimed 6403,84 credits and I have only received 3690,52 granted credits , this is just a average off 57,6 %. Is this common ?. I find it quite low.

IMHO, 25.28 CREDITS/WU seems about right to me. I figure the average wu is worth about 25 credits. The credits you claimed of 44.16 CREDITS/WU seems a bit high to me. Or maybe I am expecting too little :D

If anyone has an opinion on how many credits an average WU is worth, I'd be glad to hear it.

-Mark

In an earlier thread we had decided that it was in the low 30's... The average seems to have gone down since then though. You can see that thread here: http://setiweb.ssl.berkeley.edu/forum_thread.php?id=5889

My only guess as to why would be either changes to the benchmarking system or else higher usage of opitmized clients (which tend to request lower credit)

---

-Aardvark

But I do think a more accurate benchmarking method is needed.

Have you seen 4.73 or 5.1.x?

It pretty much has the problem solved....

---

Even with "normal" clients faster computers do not claim as much credit as they should claim. And as the computers are getting faster and faster the average claimed credit gets lower and lower. And as the average claimed credit is decreasing, the average granted credit is also decreasing.

But this also means: It's not true that you get same credit for the same work. As time goes by you get less credit for the same work...

It's called inflation.

---

Kitty@SETI team (Russia). Our cats also want to know if there is ETI out there

I sure wonder if the claimed credits are really accurate... I have two boxes crunching of which one is about three times as fast in completing a result as the other, yet they have claimed just about the same amount of credit (give or take 2-3) per result.
I just changed to an optimized client and am yet to see what change that brings. V.4.45.

Or am I way off here and should it be like that?

Iiro

I sure wonder if the claimed credits are really accurate...

This one is easy to answer. It isn't accurate. Nor is it likely to get significantly better in the near term.

There are fixes in the latest generation (4.72 and later) versions. But, these are not of release quality (whatever that means), so, we will not see a change anytime soon. Worse, even with improvements, with people using some very old versions, some of the improvements will not happen till we get them weaned off of the 4.25 and earlier ...

---

I just changed to an optimized client and am yet to see what change that brings.

Quite a change!

My slower machine, a Pentium III@800MHz dropped off a sweet 13-15 ksecs on the first completed result. Nice.

Iiro, talking to myself... A bit OffTopic of course.

It's called inflation.

Back in my day, a work unit was worth somethin'! A few honest hour o' crunchin' might earn ya 30, 40 sometimes near 50 credits fer yer troubles. But nowadays, them wippersnapper validators gettin' all cheap. No respect. No respect at all I tell ya! Why, I've got half a mind to walk right on down there and tell them validators what they can do with their precious formulas and hokus-pokus and whatnot. Can't believe it! I...

What was I talking about?

Right! Those bank tellers! I had an account with that bank for fifty-three years if it was a day, and this little snot-nosed kid wants to see my ID??? It's an outrage...

---

No animals were harmed in the making of the above post... much.

It's called inflation.

Back in my day, a work unit was worth somethin'! A few honest hour o' crunchin' might earn ya 30, 40 sometimes near 50 credits fer yer troubles. But nowadays, them wippersnapper validators gettin' all cheap. No respect. No respect at all I tell ya! Why, I've got half a mind to walk right on down there and tell them validators what they can do with their precious formulas and hokus-pokus and whatnot. Can't believe it! I...

What was I talking about?

Right! Those bank tellers! I had an account with that bank for fifty-three years if it was a day, and this little snot-nosed kid wants to see my ID??? It's an outrage...

lol

---

-Aardvark

It's called inflation.

Back in my day, a work unit was worth somethin'! A few honest hour o' crunchin' might earn ya 30, 40 sometimes near 50 credits fer yer troubles. But nowadays, them wippersnapper validators gettin' all cheap. No respect. No respect at all I tell ya! Why, I've got half a mind to walk right on down there and tell them validators what they can do with their precious formulas and hokus-pokus and whatnot. Can't believe it! I...

What was I talking about?

Right! Those bank tellers! I had an account with that bank for fifty-three years if it was a day, and this little snot-nosed kid wants to see my ID??? It's an outrage...

LOL!!!!

---

Kitty@SETI team (Russia). Our cats also want to know if there is ETI out there

A good way to find out what your average grated credit is to run AndyK's Pending Credits for BOINC.

I've been thinking that the "granted credits" box under "options" is an average but I could be sadly mistaken.

---

I've notice something that may be drawing down the averages - the last unit I just uploaded took 2 hours and 35 min to crunch, or about 9300 secs, but I was only credited ~6100 seconds, or almost a 1/3 off on the time. Maybe that is drawing down the average credit/WU?

---

IMHO it's because of the low benchmarks for P4 HT machines... very low. They crunch WU quite fast, have low benchmarks so the claimed credit for those machines are LOW....

I have a couple of AMD XP's at home & my dad has a P4 with Hyper threading.

Check out the difference in the claimed credit!

P4 HT

AMD


(edit tried to hyperlink)

---