Advanced search

Forums : General Topics : Cosmology@Home credit discussion
Message board moderation

To post messages, you must log in.

Previous · 1 . . . 7 · 8 · 9 · 10 · 11 · 12 · 13 · Next

AuthorMessage
Brian Silvers

Send message
Joined: 11 Dec 07
Posts: 420
Credit: 270,580
RAC: 0
Message 6767 - Posted: 27 Jul 2008, 0:32:00 UTC - in response to Message 6763.  
Last modified: 27 Jul 2008, 0:32:25 UTC

I know, in the past, that people have said this project gives too much credit, but to change it so that it gives even less than WCG does not seem to be the way to resolve credit issues.

That is precisely why I proposed to reward 170 credits in total, which is lower than before but still generally in the current \'middle 80%\' range for any system/OS/person.

It still gives a lot more than WCG on my system, WCG gives usually below claim

Isn\'t that an issue with WCG? Why do you demand it to be something that this or any other project has to \"resolve\" by lowering credits to match an apparent defect in another project?

I don\'t and I didn\'t. WCG is not so far off that I would ask for more credits, and it\'s a lie that Cosmo grants less than WCG.


I am not participating in WCG, so I cannot make a direct comparison, but I would tend to believe that the real answer is \"it depends\". Certain variables change what is seen by each user. Your choice of OS plays a major part in what you see as far as projects who use Benchmark * time.

In any case, BOINC-wide standings are totally meaningless because of the fact that projects are not completely even across the board in regards to Credit per CPU second. Instead of being concerned that you are losing ground in the standings BOINC-wide against others because you are not as active in projects that grant more, why don\'t you channel that energy into coming up with a real \"equal credit\" system?

BTW, yes, this means that I am sure that you are not making these types of requests out of pure altruism for BOINC as a whole. You should try to understand that the benchmark system is flawed and that forcing a flawed approach just for consistency\'s sake and so that it protects your own standings is not \"forward progress\".
ID: 6767 · Report as offensive
Profile Ananas

Send message
Joined: 19 Jan 08
Posts: 180
Credit: 2,500,290
RAC: 0
Message 6768 - Posted: 27 Jul 2008, 1:53:37 UTC
Last modified: 27 Jul 2008, 2:20:24 UTC

On WCG all subprojects I have received so far grant slightly more than claimed on 45nm CPUs under Windows.

On Cosmology some results grant slightly less than claimed on the same CPUs, but some grant more (and that \"more\" is not \"slightly\").

So I think, that in average Cosmo is now on about the same level as WCG - for those CPUs under Windows.


Comparing credits isn\'t so easy. For the efficiency (and I don\'t mean the plain CPU speed here) of a computer, the OS can play a role, compiler quality and optimisations, memory throughput, L2 cache and sure some more.

So even if two projects give exactly equal credits in average(!), it can easily happen that one box is much better for this or for that project.


Extreme examples for that are POEM and QAH. Whereas POEM really loves RAM throughput and doesn\'t care about L2 cache size at all, QAH really needs a big L2 cache, but cares not much about RAM speed.

An OCed Q9450 might beat a modern dual quadcore Xeon on POEM, whereas it has not much of a chance against the same Xeon machine on QAH.


Conclusion : It is very likely that both parties in this discussion here are right under the conditions that they have experienced. No need for a battle.


p.s.: Of course the BOINC client plays a big role too. Even 5.10.28 and 5.10.45, which aren\'t too far from eachother, seem to have totally different benchmark results on nearly identical boxes. Saenger\'s benchmark isn\'t as low as some older Linux clients would have been, I guess they have worked on that lately and brought Linux and Windows closer together.
ID: 6768 · Report as offensive
Brian Silvers

Send message
Joined: 11 Dec 07
Posts: 420
Credit: 270,580
RAC: 0
Message 6769 - Posted: 27 Jul 2008, 2:32:34 UTC

I decided to get a batch of 2.14 tasks to work on here on my AMD. I\'m currently projecting a minimum of 4 hours to complete the first task. The \"to completion\" timer is still going up though.

So, let\'s be objective...

70 / 4 = 17.5
70 / 5 = 14
70 / 6 = 11.667

Now let\'s take a look at Saenger\'s Q9450...

The worst claim/grant ratio is for a 67.9 claim.

70 / 9,646.06 = 0.0072568 cr / s
0.0072568 * 3600 = 26.124 cr / hr

If I take a look again at my host, I have a claim rate of 15.09/hr. This means that if this host takes any longer than 4.639 hours, it will be given less than the claim.

Looking at Conrad Poohs\' Pentium 4 3.00GHz system, it paints a very poor picture. As it stands, his system claimed 88.73 for 36,643.91 seconds, or only 8.717 cr / hr. He was granted 70, for a *real* cr/hr rate of 6.877. Since it is a hyper-threaded system, even doubling that gets up to a whopping 13.754, although a more realistic figure would be a 1.8x for having Hyper-Threading, so around 12.379 cr/hr.

I\'m sure my lowly Pentium 4 2.40 GHz without Hyper-Threading would be even worse, probably around 8-10 cr/hr...

Meanwhile, Saenger still thinks that this project is giving out \"too much\".

As I said, it depends on your perspective... It looks like systems with high memory bandwidth and stronger FPUs are somewhat spared, but Pentium 4 and Athlon XP systems look to be in for massive decreases in both performance and credits.
ID: 6769 · Report as offensive
Brian Silvers

Send message
Joined: 11 Dec 07
Posts: 420
Credit: 270,580
RAC: 0
Message 6770 - Posted: 27 Jul 2008, 2:44:14 UTC - in response to Message 6768.  


Conclusion : It is very likely that both parties in this discussion here are right under the conditions that they have experienced. No need for a battle.


Well, there\'s the rub. It is \"a battle\" to some. Some cannot seem to understand that their experience may be different from someone else\'s.

For example, AMD64/Windows systems had a major performance disadvantage compared to equivalent AMD64/Linux systems over at Einstein during science run S5R2. If one were to look at a Linux system, one might\'ve had the idea of \"wow, credits are too high here\", and then those of us running Windows would\'ve gotten sacked with another round of credit lowering, making other projects more attractive if even slightly interested in performance / time...

What these \"Lower Credit Crusaders\" do is just as bad as what the \"Give Me More Marauders\" do, each in their own way.

As I said earlier, I\'m sick of this pedantic behavior by both sides of the issue. Put the effort into coming up with a real plan that can provide equivalent credit across the board

or

SHUT UP ABOUT IT!

ID: 6770 · Report as offensive
Profile Ananas

Send message
Joined: 19 Jan 08
Posts: 180
Credit: 2,500,290
RAC: 0
Message 6771 - Posted: 27 Jul 2008, 3:13:20 UTC

Athlon XP 1800+ (OCed) Thoroughbread (L2=256k), CC 5.10.28 Windows :

17.04
79.49
24.68
78.96
74.67
79.00
85.52

All values are claimed and all will give me 70 credits.

I\'m not sure about the 24.68 and 17.04 yet. If those are just leftover old WUs,
the average here might really be lower than WCG - which is already known as a
low-granting project.

But if those smaller WUs always pop in now and then, it isn\'t that bad here.

We will have to watch it for a while I guess :-)
ID: 6771 · Report as offensive
Brian Silvers

Send message
Joined: 11 Dec 07
Posts: 420
Credit: 270,580
RAC: 0
Message 6772 - Posted: 27 Jul 2008, 3:43:12 UTC - in response to Message 6771.  

Athlon XP 1800+ (OCed) Thoroughbread (L2=256k), CC 5.10.28 Windows :

17.04
79.49
24.68
78.96
74.67
79.00
85.52

All values are claimed and all will give me 70 credits.

I\'m not sure about the 24.68 and 17.04 yet. If those are just leftover old WUs,
the average here might really be lower than WCG - which is already known as a
low-granting project.

But if those smaller WUs always pop in now and then, it isn\'t that bad here.

We will have to watch it for a while I guess :-)


I\'m almost positive those two will grant 50.

From what I\'ve been able to research so far by looking at the top computers list, my system will make out just barely better than claimed. The only reason for that is that it is so heavily overclocked. I have a 3700+, but it is clocked at 2.75GHz with DDR-500 memory, so the performance is somewhere between a FX-55 and a FX-57 system.

All Athlon 64 X2 3800+ systems I\'ve looked at are all claiming in the 80s on a consistent basis... A 5000+ system is claiming around 75 consistently. I haven\'t found a 5200+, which is what I\'m probably closest to, since FX systems are hard to come by...

Again, Saenger needs to step back and understand that what he is personally seeing is not what everyone else sees. There is no way I\'m bringing my Pentium 4 back to this project until something is done about the performance or the credit. I\'m borderline on my AMD too. I like the science of this project, but not if I\'m going to have to go through another round of credit lowering to suit people like Saenger...especially since that system (AMD) seems to already be right on the edge of granting less than claimed...
ID: 6772 · Report as offensive
Profile Ananas

Send message
Joined: 19 Jan 08
Posts: 180
Credit: 2,500,290
RAC: 0
Message 6773 - Posted: 27 Jul 2008, 3:46:55 UTC
Last modified: 27 Jul 2008, 3:48:15 UTC

The 24.68 (which is older than the 17.04) has already granted 70

The 17.04 has a problem, it is a re-delivered 2.12 workunit, 2.12 and 2.14 seem to be incompatible (no consensus yet)

When the next result pops in, it will most likely be a 2.14 and the WU will grant 70 - I\'m quite sure.
ID: 6773 · Report as offensive
Brian Silvers

Send message
Joined: 11 Dec 07
Posts: 420
Credit: 270,580
RAC: 0
Message 6774 - Posted: 27 Jul 2008, 4:20:19 UTC - in response to Message 6773.  
Last modified: 27 Jul 2008, 4:25:03 UTC

The 24.68 (which is older than the 17.04) has already granted 70

The 17.04 has a problem, it is a re-delivered 2.12 workunit, 2.12 and 2.14 seem to be incompatible (no consensus yet)

When the next result pops in, it will most likely be a 2.14 and the WU will grant 70 - I\'m quite sure.


Probably so. That\'s cool, I suppose, but with what you posted your average claim is 62.77. Take out those two outliers though and you\'re up to 79.53. With a grant of 70, that means the grant is about 12% less than claim.

As you said, it would depend on the frequency of the shorter-running tasks... Do you have any way of going back and making sure that lensing was turned on for those?

Edit: BTW, if those of you who have these huge farms or fast processors would stop being so secretive and not showing your computers, those of us, like myself, that have no life and don\'t mind browsing through results to find answers to these burning questions would find things a bit easier... :hint:
ID: 6774 · Report as offensive
Profile Ananas

Send message
Joined: 19 Jan 08
Posts: 180
Credit: 2,500,290
RAC: 0
Message 6775 - Posted: 27 Jul 2008, 7:11:26 UTC

No backup, so I cannot check the WU input for lensing anymore.

The redelivered ones still give 50 with 2.14 btw. ... and 0.00 for the 2.12 wingmen, even though they didn\'t have errors.
ID: 6775 · Report as offensive
STE\/E
Volunteer tester

Send message
Joined: 12 Jun 07
Posts: 375
Credit: 16,522,388
RAC: 0
Message 6776 - Posted: 27 Jul 2008, 9:38:26 UTC


Edit: BTW, if those of you who have these huge farms or fast processors would stop being so secretive and not showing your computers, those of us, like myself, that have no life and don\'t mind browsing through results to find answers to these burning questions would find things a bit easier... :hint:


Ya, I was them guys weren\'t so secretive either ... :P
ID: 6776 · Report as offensive
Profile Westsail and *Pyxey*
Avatar

Send message
Joined: 19 Dec 07
Posts: 24
Credit: 889,050
RAC: 0
Message 6777 - Posted: 27 Jul 2008, 11:33:03 UTC

mmmm kay...well now let\'s see..OK, so I was a bit bored, (not really) So decided to do as best I could figure, some math across multiple projects to see how they compare on a single host. As always YMMV.

So, these were all crunched on the same host, headless, with nothing besides the OS and Boinc running. I used a stock HP pavillion s3000. It has an AMD x2 4200+ and runs BOINC 6.3.5 under Linux: 2.6.24-19-generic.

As many WUs as I could were averaged together to calculate the results.
I still have many other WUs pending credit from some other projects I hadn\'t crunched in ahwile. However, now that I am bored, here are the results I have thus far.
In no particular order:

Cosmolgy - 17,386.77 secs earned 70 credits or 14.49 cr/hr (347.76 per day)

YOYO - 6,133.03 secs earned 37.71 credits or 22.14 cr/hr (531.36 per day)

Orbit - 866,501.05 secs earned 2,657.97 credits or 11.04 cr/hr (264.96 per day)

SETI - 5,362.2 secs earned 28.65 credits or 19.23 cr/hr (461.52 per day)

Rosetta - 10,530.74 secs earned 43.30 credits or 14.80 cr/hr (355.2 per day)

Einstein - 27,498.93 secs earned 237.53 credits or 31.096 cr/hr (746.3 per day)

LHC - 2,364.98 secs earned 6.74 credits or 10.26 cr/hr (246.24 per day)

ABC - 12,567.47 secs earned 126.72 credits or 36.29 cr/hr (870.96 per day)


ID: 6777 · Report as offensive
STE\/E
Volunteer tester

Send message
Joined: 12 Jun 07
Posts: 375
Credit: 16,522,388
RAC: 0
Message 6778 - Posted: 27 Jul 2008, 12:23:22 UTC - in response to Message 6777.  

mmmm kay...well now let\'s see..OK, so I was a bit bored, (not really) So decided to do as best I could figure, some math across multiple projects to see how they compare on a single host. As always YMMV.

So, these were all crunched on the same host, headless, with nothing besides the OS and Boinc running. I used a stock HP pavillion s3000. It has an AMD x2 4200+ and runs BOINC 6.3.5 under Linux: 2.6.24-19-generic.

As many WUs as I could were averaged together to calculate the results.
I still have many other WUs pending credit from some other projects I hadn\'t crunched in ahwile. However, now that I am bored, here are the results I have thus far.
In no particular order:

Cosmolgy - 17,386.77 secs earned 70 credits or 14.49 cr/hr (347.76 per day)

YOYO - 6,133.03 secs earned 37.71 credits or 22.14 cr/hr (531.36 per day)

Orbit - 866,501.05 secs earned 2,657.97 credits or 11.04 cr/hr (264.96 per day)

SETI - 5,362.2 secs earned 28.65 credits or 19.23 cr/hr (461.52 per day)

Rosetta - 10,530.74 secs earned 43.30 credits or 14.80 cr/hr (355.2 per day)

Einstein - 27,498.93 secs earned 237.53 credits or 31.096 cr/hr (746.3 per day)

LHC - 2,364.98 secs earned 6.74 credits or 10.26 cr/hr (246.24 per day)

ABC - 12,567.47 secs earned 126.72 credits or 36.29 cr/hr (870.96 per day)



LOL ... I laugh all the time at the ABC Project mainly because at 1 time it was one of the Worst Projects for Credit & now it\'s one of the best Projects for Credit.

A little story about the ABC Project Follows: I was a Mod there when the Project first started out, admittedly briefly though. I feel that was in part due to the fact that privately behind the scenes in e-mails I tried to explain to Hendrick that he had to raise the Credits if he ever expected the Project to ever amount to anything.

All of a sudden Hendrick said they didn\'t need any Forum Mods & they would do their own Modding of the Forums & took the Mod Job away from me and the other Mod\'s. Well the Participation at the Project started to suffer because he wouldn\'t raise the Credits like I warned him.

I went off & started running other Projects but after a time decided to check back in on the ABC Project again. Much to my surprise the Credits were way higher than what they were when I left the Project, so I don\'t know what happened other than Hendrick seen the Light I guess and raised the Credits so the Participation at the Project would raise too.
ID: 6778 · Report as offensive
Profile Saenger
Volunteer tester
Avatar

Send message
Joined: 22 May 07
Posts: 110
Credit: 282,157
RAC: 0
Message 6779 - Posted: 27 Jul 2008, 13:13:08 UTC - in response to Message 6769.  

Meanwhile, Saenger still thinks that this project is giving out \"too much\".

I don\'t and I\'ve said so. Imho they are just fine for my system. I wouldn\'t complain about 80, but would as well not complain about 60.

I\'ve seen in the other posts here that Linux obviously suits this project better so I will get more than average here, and since Windows is still the main OS that\'s fine with me. So if that\'s the reason I will not complain about even more. The same is over @Einstein, where I get the same credits per WU as the windows computers, but need only half the time. The Win app should be made as fast as mine asap, and credits stay for them i.e. reduced for me, but that\'s another project, not this one.

As you\'ve shown your c/h ratios, here are mine:


Superlink has a credit cap, I broke it quite severe several times.

My computer is this one.
Grüße vom Sänger
ID: 6779 · Report as offensive
STE\/E
Volunteer tester

Send message
Joined: 12 Jun 07
Posts: 375
Credit: 16,522,388
RAC: 0
Message 6780 - Posted: 27 Jul 2008, 13:38:06 UTC

Looking at Conrad Poohs\' Pentium 4 3.00GHz system, it paints a very poor picture. As it stands, his system claimed 88.73 for 36,643.91 seconds, or only 8.717 cr / hr. He was granted 70, for a *real* cr/hr rate of 6.877. Since it is a hyper-threaded system, even doubling that gets up to a whopping 13.754, although a more realistic figure would be a 1.8x for having Hyper-Threading, so around 12.379 cr/hr.

I\'m sure my lowly Pentium 4 2.40 GHz without Hyper-Threading would be even worse, probably around 8-10 cr/hr...


The Original Pentium 4\'s never were really any good for Crunching the Projects & you would be hard pressed to find any Project that gives them a decent amount of Credit, even the slightly newer D Model P4\'s weren\'t much better. I Upgraded all of mine as fast as I could to the Core2\'s & Quad CPU\'s when they came out.

I do have 1 P4 Laptop left though that cost me around $3200 & needs about $200-$300 to get running again but I haven\'t bothered with it because I\'m not sure I even want to spend that kind of money on it for what it will produce @ the projects.

My thinking is to just scrape it out to the junk pile & spend the $200-$300 on some of the parts to build another Quad Core Box ... :P
ID: 6780 · Report as offensive
Profile Jord
Volunteer moderator
Volunteer tester
Avatar

Send message
Joined: 15 Jun 07
Posts: 345
Credit: 50,500
RAC: 0
Message 6781 - Posted: 27 Jul 2008, 14:51:17 UTC - in response to Message 6777.  
Last modified: 27 Jul 2008, 15:02:12 UTC

Direct compare against one of my computers, a P4 3.0GHz HT, WinXP Pro SP2:
Cosmolgy - 17,386.77 secs earned 70 credits or 14.49 cr/hr (347.76 per day)

SETI - 5,362.2 secs earned 28.65 credits or 19.23 cr/hr (461.52 per day)

Einstein - 27,498.93 secs earned 237.53 credits or 31.096 cr/hr (746.3 per day)

Cosmo - 10,634.45 earned 50.00 credits or 16.93 cr/hr (406.23 per day)

SETI long - 10,502.11 earned 50.74 credits or 17.39 cr/hr (417.43 per day)
SETI short - 4,432.52 earned 20.09 credits or 16.31 cr/hr (391.42 per day)

Using Optimized application - Version info: SSE2x (AMD/Intel, Core 2-optimized v8-nographics) V5.13 by Alex Kan.

Einstein - 50,367.49 earned 237.81 credits or 16.997 cr/hr (407.937 per day)
Einstein - 53,494.74 earned 237.81 credits or 16.00 cr/hr (384.09 per day)

The thing with Einstein is that the time you spend on a task fluctuates highly, but the credit is always in the same ballpark due to them using the floating point capability of your CPU.

Just showing you mine for comparison. I\'m sure it will show differently on everyone else\'s computers. I\'m even sure it will show different outcomes on my other computer.

For all I care all projects follow QCN Alpha\'s example: no credit.
ID: 6781 · Report as offensive
Profile Copycat-Digital for WCG*
Avatar

Send message
Joined: 25 Sep 07
Posts: 17
Credit: 1,471,530
RAC: 0
Message 6782 - Posted: 27 Jul 2008, 21:02:13 UTC

Let’s face the facts:

Motor vehicles run on fuel.
Computers run on electrical power.
Human beings run on food.
Crunchers run on credits.

I agree 100% with Slicker’s post in “Why do you participate in Cosmology”

“I participate to further human knowledge of the universe. NOT!!! I crunch strictly for the credit.
Credit is all I get for the hundreds of dollars a year spent on electricity to power (and cool) my machines. So, when Cosmo drops from the top 3 BOINC projects for awarding credit, you will likely see me leave unless someone volunteers to subsidize my electric bills.”

Thanks for your honesty!

Next- Q
Why the big fluctuation in time for a w/u?

Wu_ ID 4912708 = 17059 secs. to complete
Wu_ ID 4916025 = 5851 secs. to complete
Both granted 70 credits on the same host
ID: 6782 · Report as offensive
Brian Silvers

Send message
Joined: 11 Dec 07
Posts: 420
Credit: 270,580
RAC: 0
Message 6785 - Posted: 28 Jul 2008, 5:40:56 UTC - in response to Message 6779.  
Last modified: 28 Jul 2008, 5:41:25 UTC

Meanwhile, Saenger still thinks that this project is giving out \"too much\".

I don\'t and I\'ve said so. Imho they are just fine for my system. I wouldn\'t complain about 80, but would as well not complain about 60.


Well, I\'m \"complaining\" now about 70. Turns out I was wrong. So far, both results that I\'ve reported and the one I\'m currently working on are taking over 5 hours, thus I\'m requesting 75-80, but will be getting only 70.

10866964 4958581 27 Jul 2008 0:38:34 UTC 28 Jul 2008 5:30:37 UTC Over Success Done 18,945.25 78.90 pending
10866961 4958580 27 Jul 2008 0:38:34 UTC 28 Jul 2008 3:02:41 UTC Over Success Done 18,283.63 76.15 pending


So, you tell me, is this \"fair\"? Are you at least beginning to see that what you see on your system is not necessarily what the rest of us see on ours?

This fixed credit is simply not going to work, if it is less than claimed. Either it needs to be bumped up to around 100 (for minimalists, like yourself), or more like 150. Only other thing to do is to switch to where it actually grants the claimed amount.
ID: 6785 · Report as offensive
Profile Conan
Avatar

Send message
Joined: 28 Aug 07
Posts: 169
Credit: 1,280,875
RAC: 0
Message 6791 - Posted: 28 Jul 2008, 12:20:55 UTC
Last modified: 28 Jul 2008, 12:22:23 UTC

All your views have been heard as I am now getting ZERO credit for my current 6 HOUR work units.

So the people who say \"I am only here for the science\" will be happy, as it is now only science.

All the others who cried that Cosmology pays too much credit should also be happy, as it is not paying too much any more.
ID: 6791 · Report as offensive
Profile kevint

Send message
Joined: 30 Aug 07
Posts: 46
Credit: 6,502,980
RAC: 0
Message 6796 - Posted: 28 Jul 2008, 17:14:45 UTC
Last modified: 28 Jul 2008, 17:17:27 UTC

Snedegar.

No wonder you are crazy, you are using the claimed/granted in BOINC to justify your claims of \"too much credit\"
This does not work. Neither does cross project parity.

If an employer(the project) wants to hire and retain workers(the participants) he will pay more. If he does not want to hire and retain workers, then he pays less.

So it seems that Cosmo has decided that they have enough participants, and those participants who know their real value will start looking for new places to work.

Isn\'t the free market a wonderful thing!!!!


However this ==== >>>> just does not work for anyone unless you consider yourself a slave http://www.cosmologyathome.org/workunit.php?wuid=4889562
Zero credit granted for the entire group!! - and this is not the only instance.
ID: 6796 · Report as offensive
Brian Silvers

Send message
Joined: 11 Dec 07
Posts: 420
Credit: 270,580
RAC: 0
Message 6797 - Posted: 28 Jul 2008, 17:24:41 UTC
Last modified: 28 Jul 2008, 17:25:52 UTC

Update:

During the course of the past 12 hours, I did have one task that was a short-running task. I previewed all of the .INI files ahead of time and all of them were do_lensing = T, so it was not expected. Even still, I am only just barely averaging a claim that is close to what should be granted.

Here\'s the math:

78.90
78.76
77.36
76.15
40.94

Avg. Claimed Credits: 70.422
Avg. Claimed Credits without the outlier: 77.793

It should be noted that my system is among the fastest single-core systems around. This means that AMD K7 / K8 systems are most likely all going to be claiming more than they are granted.
ID: 6797 · Report as offensive
Previous · 1 . . . 7 · 8 · 9 · 10 · 11 · 12 · 13 · Next

Forums : General Topics : Cosmology@Home credit discussion