I am all out of cpu work.

---

My 2500K ran out of SETI work this morning, but it has picked up another 4 tasks since then and my 3570K still has 45 tasks, though they'll have to hang around for a bit while I get some other tasks out of the way 1st.

Cheers.

My i72600k has been out of work for several days now.
My E6600 has actually picked up CPU work several times- mostly because it is so slow- every 4 or so hours it resets the manager backoffs.
Unfortunately the backoffs with the v7 manager are so ridiculous that there is no chance of getting any GPU work, or CPU work on the i7 system.

---

Grant
Darwin NT

I have been crunching SETI since 1999 and I will stick with SETI, I will pick up Einstein work as my machines run out. I pick up a bit of Einstein now and again anyway. This is about the driest period I can remember, so far the only machine that has "run out" is a new machine I just put online to replace some newer older machines. Now might be a good time for me to switch one of my machines to Linux as I have been planning to do for some time now. I still have work on 4 machines, I stopped one of them from getting new tasks because I'm going to stop it from crunching away. I have come to the conclusion its just not a good idea to crunch on laptops especially their GPU's. Hope everyone gets tasks soon.

---

My backup project is building another machine. i5-4670k
on a gigabyte Z97X. Case is here - mobo, cpu next week.

It's all about the science.

Will this ever gonna work again?
Don't let the volunteers crunch too much on their backup projects, f. e. Einstein, etc.
When they see the credit given there, they might give up on this... :þ

True, when people get used to the extremely much higher credits of their backup project, they will abandon SETI in droves.

That'll of course benefit those of us who stays. More work units for us, and much easier to get them, once the project starts splitting work again :-)

I will abandon my backup projects as soon as an abundant supply of owrk units from Seti@Home arrive.
Milkyway is paying credits out bigtime for two of my machines. But my other one running Einstien is meager. But then Im only crunching on cpu. I think its because Im running a single precision GTX 550 Ti card. I will have to look into my account settings and the website and see whats up.
Finally went over the 1 million mark for Milkyway. Not bad for a backup project.

---

[/quote]

Old James

Milkyway is paying credits out bigtime for two of my machines. But my other one running Einstien is meager.

Einstein is not a great soure of credits but it does very good science.

---

A great source of credits would actually be Primegrid, my 2 GPU's average around 800k a day.

---

Milkyway is paying credits out bigtime for two of my machines. But my other one running Einstien is meager.

Einstein is not a great soure of credits but it does very good science.

+1

Well like I said when Seti is back up they get NNT set untill needed again.
I want to find THE Signal.

---

[/quote]

Old James

Well, since everybody took on my phrase so well, i'll state another reason, why the credit system is better in other projects:

Over at Einstein, there are WUs with 62.5, 1000, 3333, etc. credits per WU. The best part is, the credits are always the same for a particular WU, regardless of the WU running in 20 Minutes on a very fast machine or running for 24 hours on a very slow machine. This is how a credit system should work: Same work done = same credit given. How simple is that. This holy magic formula used to "generate" credits in Seti is just insane... In reality it is just a sadly complicated random number generator, nothing more.

...but old habits are hard to brake.

So crunch on!

...well, if you can manage to get any WUs. >;)

---

Aloha, Uli

Einstein's tasks are very deterministic - it is possible to predict how much computing is required to complete a task, therefore how many credits can be awarded. SETI tasks are far from deterministic, while it is broadly possible to say "this task will take more effort than that task" this is not always the case, problems with in-coming data quality and terrestrial RFI (and more) make it a bit of a gamble as to exactly how much effort is required, and so how much credit should be awarded (and this laying aside the arguments about processor types...).

---

Bob Smith
Member of Seti PIPPS (Pluto is a Planet Protest Society)
Somewhere in the (un)known Universe?

Yes, SETI is much less deterministic.

But all of us know that we have "shorties" and longer tasks.
In few words, we learn to know in few runs how long it takes approximately (but quite precisely) for each class of WUs in our configurations.
And there are not even so many classes of WUs for MB and AP.

Therefore, it should not be very difficult to make the same assessment almost auto-magically when a new project version starts and give to each class the same credit, roughly comparable to what other projects give.

Fluctuations are very small and much smaller than the randomness and punishing stinginess of CreditNew.

Outliers are limited.
AP 6 seconds WUs are limited as well and would not destabilise the system too much. On the other hand, even CreditNew at times gives a bazillion credits for a 6 seconds AP task.

In any case, already either other projects have managed to escape CreditNew and give way too many credits (look at my combined stats of these days compared to the previous and see what I mean...) or SETI is giving too few.
Therefore, already CreditNew both because of technical reasons and "political" reasons is not working as it should, as we already know.

As we say, the best optimum is enemy of the good.

Sleepy

The unpredictability of the Seti WUs is really no argument for me, because in the end of a WU all applications know exactly how much "flops" the crunching did cost and they write it into their stdout.txt, which is sent to the server. So it's only a matter of a simple factor, to allocate the matching credit for a WU. Really, everything else is beating about the bush...

---

Aloha, Uli

http://en.wikipedia.org/wiki/BOINC_Credit_System

This is the way the boinc credit system is designed.

In short, if a computer does 1 Gflop/s it wil give you 200 cobblestones per day. Giving a fixed number of cobblestones to a Wu is possible if you know exactly how much flop's a Wu takes.

To be fair, as seti is the first project which used boinc and the new program's were developed with each other, the seti project is one of the few who's giving the correct amount of cobblestones.

The reason some projects give more is because of the people who do this solely for the credit's will go to the project which give's the most per time spent.

The number of credit's a machine can gather is therefore project specific and that's the reason I only compare machine's with each other which are doing the same project.

It's the stats sites who are making a mess of it adding the credits from multiple project's and comparing those. This could only work if all the credits are calculated the same.

---

http://en.wikipedia.org/wiki/BOINC_Credit_System

This is the way the boinc credit system is designed.

In short, if a computer does 1 Gflop/s it wil give you 200 cobblestones per day. Giving a fixed number of cobblestones to a Wu is possible if you know exactly how much flop's a Wu takes.

To be fair, as seti is the first project which used boinc and the new program's were developed with each other, the seti project is one of the few who's giving the correct amount of cobblestones.

The reason some projects give more is because of the people who do this solely for the credit's will go to the project which give's the most per time spent.

The number of credit's a machine can gather is therefore project specific and that's the reason I only compare machine's with each other which are doing the same project.

It's the stats sites who are making a mess of it adding the credits from multiple project's and comparing those. This could only work if all the credits are calculated the same.

Credit allocation is actually based on this now.
http://boinc.berkeley.edu/trac/wiki/CreditNew
also see
http://en.wikipedia.org/wiki/Random_number_generation

---

SETI@home classic workunits: 93,865 CPU time: 863,447 hours
Join the [url=http://tinyurl.com/8y46zvu]BP6/VP6 User Group[

http://en.wikipedia.org/wiki/BOINC_Credit_System

This is the way the boinc credit system is designed.

In short, if a computer does 1 Gflop/s it wil give you 200 cobblestones per day. Giving a fixed number of cobblestones to a Wu is possible if you know exactly how much flop's a Wu takes.

To be fair, as seti is the first project which used boinc and the new program's were developed with each other, the seti project is one of the few who's giving the correct amount of cobblestones.

The reason some projects give more is because of the people who do this solely for the credit's will go to the project which give's the most per time spent.

The number of credit's a machine can gather is therefore project specific and that's the reason I only compare machine's with each other which are doing the same project.

It's the stats sites who are making a mess of it adding the credits from multiple project's and comparing those. This could only work if all the credits are calculated the same.

Credit allocation is actually based on this now.
http://boinc.berkeley.edu/trac/wiki/CreditNew

That doesn't change the basic definition of the cobblestone: in theory, the average credit should have remained unchanged (theory and practice sometimes differ....). What it has done is to drive other project administrators to roll their own, with even less adherence to the formal specification. Most project administrators have proved to be too interested in their own project's science, or too short of time, to spend precious time measuring and reporting the number of FLOPS needed to complete their tasks.

Well, since everybody took on my phrase so well, i'll state another reason, why the credit system is better in other projects:

Over at Einstein, there are WUs with 62.5, 1000, 3333, etc. credits per WU. The best part is, the credits are always the same for a particular WU, regardless of the WU running in 20 Minutes on a very fast machine or running for 24 hours on a very slow machine. This is how a credit system should work: Same work done = same credit given. How simple is that. This holy magic formula used to "generate" credits in Seti is just insane... In reality it is just a sadly complicated random number generator, nothing more.

...but old habits are hard to brake.

So crunch on!

...well, if you can manage to get any WUs. >;)

and ...

Therefore, already CreditNew both because of technical reasons and "political" reasons is not working as it should, as we already know.

But they are too embarrassed to fix it because it means admitting they got it wrong.

yep, as I have indicated before, this project is knackered, well and truly and it is only going to get worse unless there is a major rethink on the project and the people running it.

---

oh dear, here we go again ...

Does anyone know if "GBT" will be S@H III along with MB and AP??

If so ... help us all ...

I can't help but believe that S@H needs to be broken into 3 separate projects .... not visible one from another. (we are down now ...this is a good time to do it and reduce the size of associated DB's)

The problem with having 3 S@H's is - we will then have a 3rd voice in the everlasting cobblestone wars .... What a waste of ink!!

3 S@H's is my vote.

Ed F

I can't help but believe that S@H needs to be broken into 3 separate projects .... not visible one from another.

Astropulse is a addition to Seti@home, Seti@home searches for narrow band signals over a longer period and astropulse is searching for shorter pulses but with a much broader band.

(we are down now ...this is a good time to do it and reduce the size of associated DB's)

There are multiple databases at seti@home, but for 3 of them we feel the effect directly if they get stuck.

from the stats page:

BOINC master database: The mysql database that contains all BOINC related information (user stats, forum messages, basic workunit/result information, etc.).
BOINC replica database: A back-up server which contains an identical copy of everything in the BOINC database. Read-only queries can be aimed at this server to lessen the load on the BOINC database.
SETI@home science database: The informix database that contains final science products returned by SETI@home clients.
Astropulse science database: The informix database that contains final science products returned by Astropulse clients.

Splitting the separate programs into it's own project also means that we run out of work even faster, also the staff than need to maintain 2(or 3) projects. Which is always more work than 3 as one.
Also that we don't have more work to crunch doesn't mean that the staff isn't doing something. They are doing some scientific "stuff" over there

---

Server Status page - Warning: number_format() expects parameter 1 to be double, string given in /disks/carolyn/b/home/boincadm/projects/sah/html/seti_boinc_html/sah_status.php on line 417

---