I've been looking recently at how credits compare across projects to see if our credit ratio needs adjusting as some have suggested. Its a little difficult to do because of result purging.

Here is how some of my machines score...
The columns are machine, and granted credits per cpu hour for the various projects. "--" means no results were found for that machine on that project.

Machine            SAH   BETA  EAH   Pred  CPDN   Ros
-------------------------------------------------------
2x2.0GHz Opteron  10.74   --    --    --   15.13   --
2x1.8GHz Opteron   8.47   --   6.05  7.35  23.38   --
2x1.8GHz PPC G5    8.35  9.36  9.52  7.57  13.62   --
1.6 GHz P4-M       7.71  8.81   --  10.90  10.56   --
2.4 GHz P4         7.70  7.87   --   8.21  17.22  8.21
3.0 GHz P4-HT      6.66   --    --    --   15.57   --
0.6 GHz PIII        --   1.46   --   2.04    --   2.95
2x0.45 GHz PII      --    --    --   2.39   1.82  2.20
--------------------------------------------------------
Apprx Ratio to SAH 1.00  1.10  0.92  1.15   1.38  1.07

So the average among those four projects (not including beta) is to give about 12% more credit than SETI@home. If you take out climate prediction (which definitely gives way too much credit) that drops to 5%.

If anyone wants to put up their own statistics (including from other projects), feel free. It will help improve my estimate. But please only a stock BOINC core client and stock applications. Using optimized core clients or applications would throw off the results.

---

@SETIEric@qoto.org (Mastodon)

I've been looking recently at how credits compare across projects to see if our credit ratio needs adjusting as some have suggested. Its a little difficult to do because of result purging.

Here is how some of my machines score...
The columns are machine, and granted credits per cpu hour for the various projects. "--" means no results were found for that machine on that project.

Machine            SAH   BETA  EAH   Pred  CPDN   Ros
-------------------------------------------------------
2x2.0GHz Opteron  10.74   --    --    --   15.13   --
2x1.8GHz Opteron   8.47   --   6.05  7.35  23.38   --
2x1.8GHz PPC G5    8.35  9.36  9.52  7.57  13.62   --
1.6 GHz P4-M       7.71  8.81   --  10.90  10.56   --
2.4 GHz P4         7.70  7.87   --   8.21  17.22  8.21
3.0 GHz P4-HT      6.66   --    --    --   15.57   --
0.6 GHz PIII        --   1.46   --   2.04    --   2.95
2x0.45 GHz PII      --    --    --   2.39   1.82  2.20
--------------------------------------------------------
Apprx Ratio to SAH 1.00  1.10  0.92  1.15   1.38  1.07

So the average among those four projects (not including beta) is to give about 12% more credit than SETI@home. If you take out climate prediction (which definitely gives way too much credit) that drops to 5%.

If anyone wants to put up their own statistics (including from other projects), feel free. It will help improve my estimate. But please only a stock BOINC core client and stock applications. Using optimized core clients or applications would throw off the results.

Cool table,

but what about the other projects do they really care about a common playgound for credits ?

Never seen or heard about one of the other projects crying for that.

So i assume you want to force them to go along with flop counting and common cross project credits( the way seti likes it) and other stuff (crap?) like the silly client shedulers, EDF, the boinc curl bug which doesn't work correctly with proxies (no fix at all,but reported a long time ago), useless debits (which are causing lot of problems filling the caches),binaries released from the beta project (which do not work corectly after all, although lot of people said it wasn't read for mainstream...) and so on ...

I'm not sure if that's the right way to go.... :-(

But might be the wrong ... corect me please ;)

---

Join BOINC United now!

but what about the other projects do they really care about a common playgound for credits ?

Never seen or heard about one of the other projects crying for that.

So i assume you want to force them to go along with flop counting and common cross project credits( the way seti likes it) and other stuff (crap?) like the silly client shedulers, EDF, the boinc curl bug which doesn't work correctly with proxies (no fix at all,but reported a long time ago), useless debits (which are causing lot of problems filling the caches),binaries released from the beta project (which do not work corectly after all, although lot of people said it wasn't read for mainstream...) and so on ...

I'm not sure if that's the right way to go.... :-(

But might be the wrong ... corect me please ;)

Of course the other side of this is:

Do we want crunchers to "shop" for credits -- should SETI be able to draw people from other projects by offering "double credit"?

... or the other way around?

Should we allow other projects to hijack machines by "gaming" for time based by deadline (I'm pretty sure a project could do this by setting short deadlines and giving overlong estimates of completion time -- forcing EDF)?

---

Of course the other side of this is:

Do we want crunchers to "shop" for credits -- should SETI be able to draw people from other projects by offering "double credit"?

... or the other way around?

Should we allow other projects to hijack machines by "gaming" for time based by deadline (I'm pretty sure a project could do this by setting short deadlines and giving overlong estimates of completion time -- forcing EDF)?

It's not about "shopping" it's about the new way of credit claiming that S@H instroduced itself to "level the playground" BUT i don't see any of the other projects asking for a leveling. Most of them don't even care!

So the only conclusion left for me is that seti wan't to enforce them doing the same way...

P.S.
About the EDF mode, well it's one of the examples i took from the SETI.Germany Board.

There were a discussion about those kind of crap... but none of them posted their concerns here.

Another discussion over at the SETI.Germany board was to stop processing S@H work because of the WU times and the "leveled playground" and 98% of SETI.Germany crunchers don't use any optimized boinc or seti app.

But hey its only 10% difference after all... wasn't it that way ? look at the tables for yourself http://setiathome.berkeley.edu/top_teams.php... from were SETI.Germany has dropped to ...( remember 98 % of them are using the stock app. and NO the outage yesterday didn't do any harm at all)...


Man i'm really p****d of ...
Can't take this crap any more ...

This whole c++ bug, timer crap could have been resolved a long time ago ... but ignoring people doesn't help getting things solved..

---

Join BOINC United now!

I’ve done a quick survey of what my Mac G4/733 (BOINC Menubar 5.2.13) has been up to recently. I’m noting both claimed and granted credit, because in the other projects, due (I believe) to differences in the type of computation between the benchmarks and the science apps, Macs tend to underclaim quite considerably. OTOH one sometimes gets matched with optimized hosts that claim much less than standard ones. So IME credit grants are something of a lottery, to which a small sample probably doesn't do justice. I'll try and collect more results over the next while and report back, but in the meantime here’s a taste:

Application   Claimed/h  Ratio  Granted/h  Ratio
S@h Enh. v5.13   3.186   1.000     3.181   1.000
Einstein v4.56   2.895   0.909     5.879   1.848
SZTAKI   v1.12   3.085   0.968     4.182   1.315

Some of the results in the sample are still pending, so the grant averages are somewhat weaker than those for claims. Note also that none of the E@h results were granted less than claimed, which is certainly not always the case, and that none from SDG came up invalid—also a little unusual, although the admin often grants credit manually for such WUs after the fact. Also worth mentioning is that the variance in SDG CPU-times (and therefore claims) is proportionally much greater than that from Einstein.

---

I agree with Crunch3r, regarding SETI Enhanced credit vs. other projects. SETI was here first, was it not? Why not let the other projects adapt to SETI's credit system, if it's that important for them to draw people away from SETI?

Whenever the credit granting system is altered to the extent it was with enhanced, it makes existing stats totals meaningless, IMO. It also makes it extremely unfair for the newer up-and-coming teams and individuals to catch up to the others' stats, because the established participants were gaining credit at twice the current rate under the previous credit system.

I also feel that a lot of issues have been largely ignored by those who have the decision-making authority to correct them.

I share Crunch3r's frustration completely!

---

http://www.setiusa.net

Of course the other side of this is:

Do we want crunchers to "shop" for credits -- should SETI be able to draw people from other projects by offering "double credit"?

... or the other way around?

Should we allow other projects to hijack machines by "gaming" for time based by deadline (I'm pretty sure a project could do this by setting short deadlines and giving overlong estimates of completion time -- forcing EDF)?

It's not about "shopping" it's about the new way of credit claiming that S@H instroduced itself to "level the playground" BUT i don't see any of the other projects asking for a leveling. Most of them don't even care!

So the only conclusion left for me is that seti wan't to enforce them doing the same way...

P.S.
About the EDF mode, well it's one of the examples i took from the SETI.Germany Board.

There were a discussion about those kind of crap... but none of them posted their concerns here.

Another discussion over at the SETI.Germany board was to stop processing S@H work because of the WU times and the "leveled playground" and 98% of SETI.Germany crunchers don't use any optimized boinc or seti app.

But hey its only 10% difference after all... wasn't it that way ? look at the tables for yourself http://setiathome.berkeley.edu/top_teams.php... from were SETI.Germany has dropped to ...( remember 98 % of them are using the stock app. and NO the outage yesterday didn't do any harm at all)...


Man i'm really p****d of ...
Can't take this crap any more ...

This whole c++ bug, timer crap could have been resolved a long time ago ... but ignoring people doesn't help getting things solved..

i think Eric Korpela and some other people have to think about this....

the work from Crunch3r is extraterrestially

but what i know some people also at Berkerley are not well informed or what...

Greetings from Germany NRW
Ulli

While receiving credit and watching team stats and such is fun, the only credit that really means anything is the credit someone will get when their WU turns out to be the one from our little green friends.

I've run crunch3r's Op but only because My machines are old and slow. Optimizing allowed me to get the most out of them. I still don't quite understand the credit bit but I thought it was just to show how well we were doing.( I do still want a toaster though :D )

---

PROUD MEMBER OF Team Starfire World BOINC

Let's not make this thread yet another "My RAC is falling, ow you bad Seti!". If you want to continue posting about that, then use this thread for that.

Also, I just deleted two posts in this thread. I will not have old flamebaits/flamethreads be brought up from the past. If you call that censoring, then so be it.

So far I've seen one person give his own statistics in answer to Eric. That's what this thread is for. Not your continued whining on what is wrong with Seti Enhanced or what errors it produces or that Crunch3r's version is better. There's other threads for those kinds of comments.

Please behave.

---

My Post was 'moderated' and i'm not going to fight against windmills. I thought i was free, silly me and sorry for my 'deviating' opinion.

---

Aloha, Uli

Ok, an a related note:

Tony where are you?

Get back on board, you've always been contributing a big part to this forum!


Regards Hans

---

So the only conclusion left for me is that seti wan't to enforce them doing the same way...

I don't care about enforcing this method of doing things. I do care about not angering other projects. It's called being considerate.

Another discussion over at the SETI.Germany board was to stop processing S@H work because of the WU times and the "leveled playground"...

That would definitely hurt the team's RAC.

But hey its only 10% difference after all... wasn't it that way ? look at the tables for yourself http://setiathome.berkeley.edu/top_teams.php... from were SETI.Germany has dropped to...

First in total credit, second in RAC. Where were they before? (Personally, I have always felt that total credit should be the default sorting mechanism and that RAC should be the credit in the last 30 days/30 rather than some voodoo exponential average that behaves in a weird manner when crunch times get long and really weird manner when crunch times vary. As currently defined RAC is just too variable.).

This whole c++ bug, timer crap could have been resolved a long time ago ... but ignoring people doesn't help getting things solved.

I'm not ignoring people, but I am just one person who can only do so much in an 11 to 14 hour work day. If I work on the application, I get blamed for ignoring the message boards. And vice versa. Finding and squashing the bugs without introducing new ones isn't a cake walk.

But I really did start this thread to get data for adjusting the credit ratio. If you've got complaints about how much credit you are getting, give me the data I need to fix it. If you want to argue, please move to a different thread.

---

@SETIEric@qoto.org (Mastodon)

My small sample here for this host@seti host@rosetta Its an AMD XP1900+ no overclocking. I used the last 6 results for my stats. Also using CC 5.4.9 on Win XP Pro sp2.

So here are my results
S@H 5.77 cr/h
Rosetta 8.31 cr/h

For a difference of about 44%. So either S@H is under or Rosetta is over.

Can't compare to Einstein as I'm running optimized apps on those. My two other boxes are running optimized enhanced apps. So this is all I have.

---

Yeah, rosetta is a tricky one. No validation against other results, so what you claim is what you get. I expect it will vary greatly between hosts. We'll need lots of rosetta statistics to get a good idea of what the average is.

Thanks, it's another point to add to my calculation.

Eric

My small sample here for this host@seti host@rosetta Its an AMD XP1900+ no overclocking. I used the last 6 results for my stats. Also using CC 5.4.9 on Win XP Pro sp2.

So here are my results
S@H 5.77 cr/h
Rosetta 8.31 cr/h

For a difference of about 44%. So either S@H is under or Rosetta is over.

Can't compare to Einstein as I'm running optimized apps on those. My two other boxes are running optimized enhanced apps. So this is all I have.

---

@SETIEric@qoto.org (Mastodon)

Edit- moved response to Steve Akers post to a more appropriate thread.

---

Grant
Darwin NT

Eric:

Here is some data I have put together from various computers, BOINC clients and science apps.

Seti at Home:

Genuine Intel (R) Pentium(R) 4 650 CPU 3.40GHz
5.3.12.tx36 BOINC client and standard Seti V4.18 (average of 23 work units)
CPU time = 2,072.61 sec., Claimed credit = 16.81 CS/hr., Granted credit = 23.30 CS/hr.

5.3.12.tx36 BOINC client and Crunch3r’s Seti SSE3 V4.11 (average of 97 work units)
CPU time = 3,110.58 sec., Claimed credit = 34.90 CS/hr., Granted credit = 29.81 CS/hr.

5.3.12.tx36 BOINC client and Crunch3r’s Seti SSE3 V5.11 (average of 86 work units)
CPU time = 10,213.65 sec., Claimed credit = 46.27 CS/hr., Granted credit = 17.81 CS/hr.

5.3.12.tx36 BOINC client and Crunch3r’s Seti SSE3 V5.12 (average of 31 work units)
CPU time = 13,751.11 sec., Claimed credit = 14.38 CS/hr., Granted credit = 14.13 CS/hr.

Authentic AMD Athlon(tm) 64 Processor 3000+
5.3.12.tx36 BOINC client and Crunch3r’s Seti SSE3 V5.12 (average of 11 work units)
CPU time = 9,033.28 sec., Claimed credit = 13.56 CS/hr., Granted credit = 13.93 CS/hr.

Authentic AMD Athlon (tm) 64 Processor 3000+
5.3.12.tx36 BOINC client and Crunch3r’s Seti SSE3 V5.11 (average of 26 work units)
CPU time = 11,120.90 sec., Claimed credit = 53.18 CS/hr., Granted credit = 20.08 CS/hr.

Einstein at Home:

Intel(R) Pentium(R) D CPU 2.66GHz
5.4.9 BOINC and U41.04 akosf (average of 34 work units)
CPU time = 2,479.36 sec., Claimed credit = 71.80 CS/hr., Granted credit = 61.37 CS/hr.

Intel(R) Pentium(R) 4 CPU 3.20GHz
5.4.9 BOINC and S41.07 akosf (average of 50 work units)
CPU time = 3,096.05 sec., Claimed credit = 66.45 CS/hr., Granted credit = 55.71 CS/hr.

Genuine Intel Intel(R) Pentium(R) 4 CPU 3.40GHz
5.4.6 BOINC and 4.37 std. (average of 59 work units)
CPU time = 13,923.19 sec., Claimed credit = 8.65 CS/hr., Granted credit = 8.69 CS/hr.

Regards
Franz

Eric,

Just a suggestion:

Could you set the DB_Purger to allow the completed results to hang around a little longer (maybe a month) for a while? It would make it a litte easier to grab data and do a quick "eyeball" eval, especially on "slugs".

Regards,

Alinator

Here is an update to my one host. Was able to get some LHC units to compare.

LHC 4 units, 19413.84 sec claimed 44.91 credits for 8.32 c/hr
Rosetta 6 units 256021.82 sec claimed 591.62 credits for 8.31 c/hr
Seti 5 units 166357.20 sec claimed 255.9 credits for 5.53 c/hr

A result dropped off the host list here on seti. This is a stock 5.4.9 client, no tweaks. XP 1900+. Will have more as soon as my other two have finished their optimized work and run through some non-optimized work.

@Franz
Why are you running Crunch3rs 5.11? You know that seriously overclaims on credits right?

---

---