But unless someone can show him an error in his code nothing will change.

There isn't an error with the code - it's working as designed. The error is with the design - applying a theoretical concept of efficiency that results in real world devices that are more efficient being penalised by real world devices that are less efficient.
ie a device that produces more work in a shorter period of time is considered (by the theory) to be less efficient than a device that takes longer to do the same work.
In the real world the slower device is considered less efficient. In the theoretical world based on the potential of the devices, it's not.

We're not in the real world. We're in the world of academia.
From the lab side there is nothing wrong.

Now if some one wants to come up with $100.00 USD an hour
to pay the coder in question to come up with a Creditnew
that pleases the complainers something might get done.
Until such time...

Bill, and there in lies the problem. There are two issues with what you are suggesting: the first issue is that the academics are not willing to move <colourful_language> and for all appearances have their heads in the sand </colourful_language>, the second is that you cannot go to solution mode without identifying the issue and its root cause first.

The starting point is the work units that are being completed and returned to Berkeley coupled with analysis of type, floating point operations, credit, run time, cpu time, application, device, etc (in short as much data as you can collect that is associated with each task/work unit to aid with analysis).

The data should be collected and placed in an offline database. Analysis of the data then needs to start at testing against the original intent of the logic that formed the credit system (not the code that was developed, but the intent) with observation of impact (what is actually happening).

Bill, I give you this. When I ran only "enhanced" daily credit was circa 220+/-k. When I only run AP, daily credit is circa 220+/-k. When I run only v7, daily credit is circa 110+/-k (roughly 50% less). Prima facia, doesn't this tell you that something is wrong!!

I am not the only one who sees a 50% drop in credit on v7 work units and with so many people complaining doesn't this also tell you that something is amiss!!

You can also look at it this way, when a work unit that asks for a higher level of processing(v7 wu) gets less credit than a work unit that asks for less processing then the logic of the code must be wrong. Would you not agree.

If we can all see this, then why can't "they"?

As to cost, I'm quite sure they could get a bunch of 2nd or 3rd year computer science students to do the analysis, write new code and test it for the appropriate credit.

---

Well,look at it from their side, why should they do it?

It costs them nothing to do nothing.

Edit Cost is not only in dollars. There's also time and effort.

Well,look at it from their side, why should they do it?

Because some people (not me, not you) might leave SETI for other projects to crunch and get more credits.
I know I know it is not about credits.
But for someone may be.
When I look at general (BOINC general) ranking, I see people with 3e6 RAC, who are not in SETI and who in SETI would get much less, but still would do a big fair amount of work. If THEY are after credit, they are lost for SETI.
Is this good?

We must not introduce inflation in the credit system, but here we are causing the same deflation effect that is affecting economy in the real world.
Which is not good all the same.

Happy crunching!

Sleepy
_____________________________________________________
100 g of ham is 100 g of ham and must be paid as such.

---

Well,look at it from their side, why should they do it?

It costs them nothing to do nothing.

Edit Cost is not only in dollars. There's also time and effort.

This is not scientific thinking, this is commercial thinking.
So newbies to the project could think this is a commercial project, but in reality it is a scientific project. What will the newbies do? After some crunching they will leave the project because it's not scientific!

Two conclusions: 1. unluckily they left, we now don't have their power of computation or 2. glad they left so they don't use bandwith by up-/downloading.
For SETI (including Astropulse) as a scientific Project, which one is better 1. or 2.?

Just have a look to the server status page. Every time there are AP-WUs, they're gone nearly "instantly". This is for sure cause they gain higher credits.

---

Aloha, Uli

Is just me? or anyone else remember someone who say´s... we need to wait the slowest hosts start to sending it´s WU to creditnew do it´s magic and stabilizes the credit system... that will take 4-5 weeks...

OK the 4-5 weeks passed, the slowest hosts sends their WU and... nothing happening,... creditnew stills "paid" about 40% less credit to the compelated MB tasks if compared with V6, someone must made something about that before all switch to crunch AP only and leave the MB aside, that´s allready starting to happening if you see the servers page.

---

Is just me? or anyone else remember someone who say´s... we need to wait the slowest hosts start to sending it´s WU to creditnew do it´s magic and stabilizes the credit system... that will take 4-5 weeks...

OK the 4-5 weeks passed, the slowest hosts sends their WU and... nothing happening,... creditnew stills "paid" about 40% less credit to the compelated MB tasks if compared with V6, someone must made something about that before all switch to crunch AP only and leave the MB aside, that´s allready starting to happening if you see the servers page.

Yes, it was me.

And i`m getting exactly the same amount of credits for a mid range WU as i did with V6 ~100cr.

---

With each crime and every kindness we birth our future.

Is just me? or anyone else remember someone who say´s... we need to wait the slowest hosts start to sending it´s WU to creditnew do it´s magic and stabilizes the credit system... that will take 4-5 weeks...

OK the 4-5 weeks passed, the slowest hosts sends their WU and... nothing happening,... creditnew stills "paid" about 40% less credit to the compelated MB tasks if compared with V6, someone must made something about that before all switch to crunch AP only and leave the MB aside, that´s allready starting to happening if you see the servers page.

Yes, it was me.

And i`m getting exactly the same amount of credits for a mid range WU as i did with V6 ~100cr.

But you do more crunching on the v7 WU compared to the v6 WU. It takes much longer time (so they say, I dunno since I haven't been doing any v7 other than on my slowest Celeron Lappie), so in fact even if you get the same 100cr, it's actually worth much less.

I reckon that it'll be Christmas or New Year by the time the credits balance out.

Cheers.

All I can see right now is all my five computers are still going downhill like a sled. Too much more and may just say the h#$% with this. Didn"t mine the money when I was able to keep improving but this deflating my interest in the project.

I know what it is about, as can be seen on the start date but also with a RAC status; it gives a person a sense of accomplishment or satifaction that the person is doing something.

Hope this makes sense but me is losing interest.

bj

---

Is just me? or anyone else remember someone who say´s... we need to wait the slowest hosts start to sending it´s WU to creditnew do it´s magic and stabilizes the credit system... that will take 4-5 weeks...

OK the 4-5 weeks passed, the slowest hosts sends their WU and... nothing happening,... creditnew stills "paid" about 40% less credit to the compelated MB tasks if compared with V6, someone must made something about that before all switch to crunch AP only and leave the MB aside, that´s allready starting to happening if you see the servers page.

Yes, it was me.

And i`m getting exactly the same amount of credits for a mid range WU as i did with V6 ~100cr.

So you must be a very lucky one, if you look my WU validated they are in the range of 40 - to 110 tops, most of them are in the 80-90 range. With V6 a 120 or more was easy to see.

A host that was one of the top 10 in SETI before V7 with about >100k now strugles to reach 60K. And i don´t crunch AP at that time, all come from MB crunching only!

But that back to the new/old questions, why the AP work now "paid" about 2x the credit than MB? Something must be changed and can´t blame the correlation for that change. A 553.52+172.71 sec to crunch WU receive 35.91 credit for example and a 946.62+258.09 sec receive 99.24 credit on the same host, something is clearely not working fine, just do the simple math.

That´s just confirm me, faster and more MB/GPU RAC based is your host worst is your fall.

But i still have few beers in stock so i will keep crunching anyway, just now with 50% less RAC... I´m so happy with credit new :(

---

I think the VLAR's are having an effect. Since they delayed moving VLAR's
away from NVIDIA it might take another two weeks before the total effect is seen.

I also see the average credit rising towards V6 levels hopefully they will continue to rise.

[TiC] To bring credit parity to our Universe, they seem to know how to easily lower credit per task, but raising it is another matter/energy. Just make an AstroPulse_V7 using knowledge gained from SetiatHome_V7.[/TiC]

All of the above or below (depending on how you sort these entries) is very likely to change again for Kepler tasks.

Seriusly i hope you both are right, i realy not an AP crunching fan.

---

Example processing two at a time on a GTX660
These two workunits are not VLAR's but I do not think the credit system makes a distinction whether they are or not, just that now they are not run on NVIDIA's.

Just a few weeks ago these two ar's at .307 would have been in the 90's
Now they are both over 110 the only "ONLY" difference between the two work units
is whom they are compared against.
135.37 SETI@home v7 Anonymous platform (NVIDIA GPU) Wingperson is CPU
118.97 SETI@home v7 Anonymous platform (NVIDIA GPU) Wingperson is GPU Cuda 4.2

http://setiathome.berkeley.edu/result.php?resultid=3075265260
http://setiathome.berkeley.edu/result.php?resultid=3075265258

It takes time for VLAR's to not only be evaluated on CPU's but for all the VLAR's we processed on NVIDIA's to be eliminated from the credit calcs in the system if that is even possible.

IMHO - We seemed to have an extraordinary amount of VLAR's at the beginning of V7 so many in fact that my GTX660 got stuck on Cuda 42's due to processing cuda 50's for many many VLAR's when the switchover away from NVIDIA VLAR's occurred

Was a long time i did not receive Vlars in my GPU´s after they fix that problem.

Just to make an example what i talk about: (all on the same host with 2x690 crucnhing 2WU at a time - actualy a total of 8GPU/8CPU wu simultaneusly)

WU 1 -http://setiathome.berkeley.edu/workunit.php?wuid=1280632750
WU 2 -http://setiathome.berkeley.edu/workunit.php?wuid=1280263361
WU 3 -http://setiathome.berkeley.edu/workunit.php?wuid=1280270255

Time to crunch (GPU+CPU) Credit
WU 1 - 940.96+141.09 40.42
WU 2 - 953.66+130.28 101.67
WU 3 - 1,236.75+152.02 98.59

Comparing

WU 1 with WU 2 about the same time to process and W1 receives 40% of the credit of W2

WU3 with WU2 - W3 need a lot more time to process and receive less credit than WU2

Someone say´s is because my wingmate, but that not logical, the work done on my host have nothing to do with my wingmate hosts, we all expect more processing time more credit, simple like that... but that´s is not what realy happens here with creditnew.

I agree the time to process is directly related to the WU itself, AR and other parameters, but credit must be related to the processing time used to crunch the WU itself on that host not in other host, makes little sense to me that.

As i say on other thread, 1 kilo is 1 kilo, no matter if is lead or cotton and whatever instrument used to measure, so the credit guaranted for a determinate unit must be related to the WU itself (because the AR, etc.) and not related to the hosts who crunch that WU. That´s simple and easely to understand, if you have a faster or slower CPU/GPU or you use Nvidia or ATI or Mac makes no diference, the credit of a determinate WU must be the same not related to what host is crunching it.

---

I also see the average credit rising towards V6 levels hopefully they will continue to rise.

Crunching MB only, my RAC continues to fall.

---

Grant
Darwin NT

Juan

The diff between WU1 and WU2 are the angle range looking at most of your recent
GTX690 valids The ones with runtimes greater than 800 seconds and around 50 credits all seem to have an angle range of 1.4 to 1.7

Grants tasks also show this low credits for angle ranges between 1.4 and 1.7

Why would the angle range be the deciding factor for CreditNew??

Anyone?

PS - My low credits also seem to be around 1.4 to 1.7 I just don't have as many
as Juan

Why would the angle range be the deciding factor for CreditNew??

Anyone?

PS - My low credits also seem to be around 1.4 to 1.7 I just don't have as many
as Juan

LOL - Maybe is because the Random Number Generator (technicaly called creditnew) uses this AR as a "seed".

It´s realy wierd no? If you look for the other side, as they use about the same processing time they must receive credit at least something very close not 40% as show.

I know could be because the wingman explanation, but that´s is dificult to understand in human terms....

A fixed number of credit asssing to the WU at the generating stage (that takes AR in consideration) could be a better way to make the things more "logical" and less "academic" and made the credit totaly independant of what host process it. If you have a fast host you made more WU per day and get more credit, the way the things are that clearely is not happening as show by the example.

Could be my personal point of view but the WU must receive the same credit no matter what host crunch it, that´s clearely don´t happening with creditnew.

---

Well,look at it from their side, why should they do it?

Because some people (not me, not you) might leave SETI for other projects to crunch and get more credits.

Some people have left, not enough to make a difference
in the amount of science done. You would probably need 75%+
of the crunchers to leave to get something accomplished and that's not
going to happen.

I doubt that 1% even noticed the difference in reduced credits
per work unit crunched. Kind of like telling a politician you're
not going to vote for him and a lobby telling a politician they
have 500,000 votes that aren't going to vote for him. Who do you
think he's going to pay attention to.

<cynic> The guy who just donated $500,000 dollars to his campaign;
that's who.</cynic>

I know I know it is not about credits.
But for someone may be.
When I look at general (BOINC general) ranking, I see people with 3e6 RAC, who are not in SETI and who in SETI would get much less, but still would do a big fair amount of work. If THEY are after credit, they are lost for SETI.
Is this good?

We must not introduce inflation in the credit system, but here we are causing the same deflation effect that is affecting economy in the real world.
Which is not good all the same.

Happy crunching!

Sleepy
_____________________________________________________
100 g of ham is 100 g of ham and must be paid as such.

Actually I think the whole credit thing is a poor idea.

I think something more tangible, like for every xyz work units
you would get a password to a download of a astronomical picture
suitable for desktop wallpaper/screen savers that wasn't available
to the general Internet would be better, IMO.

This shows me you simply dont know how it works.
Only AR 0.45 and lower gives average of 100 credits.
AR higher than 1 are VHARs and give less than half because they contain less science.
It also shows you are wasting time not freeing enough CPU cores.
So dont blame credit new and V7 for falling RAC.
A VHAR takes half the time of a mid range unit if processed correctly.

Check my host.

---

With each crime and every kindness we birth our future.

Well,look at it from their side, why should they do it?

It costs them nothing to do nothing.

Edit Cost is not only in dollars. There's also time and effort.

This is not scientific thinking, this is commercial thinking.

I would say it's pragmatic thinking. The scientists, no doubt, probably
have many more and much bigger fish to fry.

So newbies to the project could think this is a commercial project, but in reality it is a scientific project. What will the newbies do? After some crunching they will leave the project because it's not scientific!

Two conclusions: 1. unluckily they left, we now don't have their power of computation or 2. glad they left so they don't use bandwith by up-/downloading.
For SETI (including Astropulse) as a scientific Project, which one is better 1. or 2.?