Welcome to MilkyWay@home

credit comparison to other projects


Advanced search

Message boards : Number crunching : credit comparison to other projects
Message board moderation

To post messages, you must log in.

Previous · 1 . . . 11 · 12 · 13 · 14 · 15 · Next

AuthorMessage
ProfileSlicker [TopGun]

Send message
Joined: 20 Mar 08
Posts: 46
Credit: 69,382,802
RAC: 0
50 million credit badge14 year member badge
Message 9370 - Posted: 29 Jan 2009, 22:58:18 UTC - in response to Message 9349.  

Arion.

Based on your numbers:

MW - 31.72 credits / hour.
EH - 15.50 credits / hour.

However, remember DA wants Seti to be the standard. So how does MW compare to Seti stock, or how does Einstein stock compare to Seti stock? It is possible Einstein is a lower paying project than Seti.


What if we all agree that WCG should be the standard? Will DA get on board and tell SETI to lower their credits to WCG's level? Not. So why do hundreds of thousands of BOINC users have to do whatever DA wants? Shouldn't it be run by committee of elected representatives? Why can't Obama decide? Or me? Or you? I digress...

Using credit per cpu second generally works OK, but it isn't necessarily a true record of what is happening. It doesn't take into account the network bandwidth a project uses, the time spent uploading and downloading WUs, the time spend downloading new application versions (hint, hint, hint), etc. It also makes a difference whether a project is geared towards a specific operating system, processor type, 32bit vs 64 bit, L2 cache, floating point vs. integer calculations, etc. So, just because one machine does double on a project compared to another doesn't mean that it actually earns double the credits per day as another. Go grab some 2-5MB WUs that some projets have and download them over a modem and see how many credits you really get each day. Anyway, looking at a single machine (processor type and features, memory, disk speed, network bandwidth, operating system) does not necessarily represent the stats for all other machines.

For exmaple, ABC grants double the credit per WU for 64 bit machines because they can finish the WU in half the time as a 32 bit machine, and that is while using the STOCK application. Does DA go after ABC to lower their 64 bit credit so that it matches seti? Nope. The 32bit app is the standard. So, I agree with the comment about comparing apples to apples. Stock app to stock app.

Lastly, as far as I'm concerned, alpha and beta projects NEED to have higher credit in order for people to put up with the issues involved in getting them up and running, app errors, system down time, etc. that being said, there should probably be a limit to how long they can remain in alpha or beta status or be pulled from the stats sites.
ID: 9370 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
ProfilePaul D. Buck

Send message
Joined: 12 Apr 08
Posts: 621
Credit: 161,934,067
RAC: 0
100 million credit badge14 year member badge
Message 9373 - Posted: 30 Jan 2009, 0:08:31 UTC

ABC also awards some 1K credit tasks. I don't know what the criteria is for selection, but I have gotten several where I asked for, say, 300 and got the 1K award... Not complaining, just commenting.

For a long time the Linux boxes way under claimed on credit because of issues with the benchmarking ... I understand that has been fixed ...

But, the other comments are correct... I run WEP on my slow Linux box and on my "fast" Mac Pro ... the linux box finishes them in half the time ... why? I would guess the compiler ... same size task, yet the "slower" computer does the task faster ...

But the problem of using SaH as the standard is that they keep deflating the award over time... which is not what the original intent was of using the Cobblestone.
ID: 9373 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
ProfileGavin Shaw
Avatar

Send message
Joined: 16 Jan 08
Posts: 98
Credit: 1,371,299
RAC: 0
1 million credit badge14 year member badge
Message 9374 - Posted: 30 Jan 2009, 0:14:29 UTC - in response to Message 9370.  

Arion.

Based on your numbers:

MW - 31.72 credits / hour.
EH - 15.50 credits / hour.

However, remember DA wants Seti to be the standard. So how does MW compare to Seti stock, or how does Einstein stock compare to Seti stock? It is possible Einstein is a lower paying project than Seti.


What if we all agree that WCG should be the standard? Will DA get on board and tell SETI to lower their credits to WCG's level? Not. So why do hundreds of thousands of BOINC users have to do whatever DA wants? Shouldn't it be run by committee of elected representatives? Why can't Obama decide? Or me? Or you? I digress...

Using credit per cpu second generally works OK, but it isn't necessarily a true record of what is happening. It doesn't take into account the network bandwidth a project uses, the time spent uploading and downloading WUs, the time spend downloading new application versions (hint, hint, hint), etc. It also makes a difference whether a project is geared towards a specific operating system, processor type, 32bit vs 64 bit, L2 cache, floating point vs. integer calculations, etc. So, just because one machine does double on a project compared to another doesn't mean that it actually earns double the credits per day as another. Go grab some 2-5MB WUs that some projets have and download them over a modem and see how many credits you really get each day. Anyway, looking at a single machine (processor type and features, memory, disk speed, network bandwidth, operating system) does not necessarily represent the stats for all other machines.

For exmaple, ABC grants double the credit per WU for 64 bit machines because they can finish the WU in half the time as a 32 bit machine, and that is while using the STOCK application. Does DA go after ABC to lower their 64 bit credit so that it matches seti? Nope. The 32bit app is the standard. So, I agree with the comment about comparing apples to apples. Stock app to stock app.

Lastly, as far as I'm concerned, alpha and beta projects NEED to have higher credit in order for people to put up with the issues involved in getting them up and running, app errors, system down time, etc. that being said, there should probably be a limit to how long they can remain in alpha or beta status or be pulled from the stats sites.


Arion asked for his numbers to be checked. That's what I did.

As for credit high or low I didn't comment. I just pointed out that DA wants Seti stock as the standard, not Einstein stock.

But if we are going to have a standard why don't those that are lower get asked to bring up their 'pay rate'. Rosetta pays less than Seti and I think Rosetta is more valuable. I like to help out Rosetta, but with a low pay rate it is a killer on crunchers RAC and that puts people off. I do intend to get my Rosetta contribution up to 100,000 but that will take a lot of time.

Never surrender and never give up. In the darkest hour there is always hope.

ID: 9374 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
ProfileCori
Avatar

Send message
Joined: 27 Aug 07
Posts: 647
Credit: 27,592,547
RAC: 0
20 million credit badge15 year member badge
Message 9375 - Posted: 30 Jan 2009, 0:15:08 UTC - in response to Message 9373.  

ABC also awards some 1K credit tasks. I don't know what the criteria is for selection, but I have gotten several where I asked for, say, 300 and got the 1K award... Not complaining, just commenting...

ABC rewards 1K credits (maximum!) to the longest WUs; their credit policy is explained here. ;-)
Lovely greetings, Cori
ID: 9375 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profilebanditwolf
Avatar

Send message
Joined: 12 Nov 07
Posts: 2425
Credit: 524,164
RAC: 0
500 thousand credit badge14 year member badge
Message 9382 - Posted: 30 Jan 2009, 1:20:08 UTC - in response to Message 9374.  

Rosetta pays less than Seti and I think Rosetta is more valuable. I like to help out Rosetta, but with a low pay rate it is a killer on crunchers RAC and that puts people off. I do intend to get my Rosetta contribution up to 100,000 but that will take a lot of time.


I finally went past 71,000 this month after 3 years, mind you 1 pc & not solid 24 hours though, but it's credit has gone down since I joined. I currently get 5-9 credits/hour, usually ~6/hr. It used to be double that 2 years ago. At the time I joined Rosetta I did seti as well and it wasn't that far off. If I remeber Rosetta was 12-16/hr, Seti (stock) was 14-15/hr. That was 2 1/2-3 years ago.
Doesn't expecting the unexpected make the unexpected the expected?
If it makes sense, DON'T do it.
ID: 9382 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Misfit
Avatar

Send message
Joined: 27 Aug 07
Posts: 915
Credit: 1,503,319
RAC: 0
1 million credit badge15 year member badge
Message 9402 - Posted: 30 Jan 2009, 3:05:49 UTC

Projects are like professional sports teams. Unless the league has a salary cap you can play the players whatever they want. So unless DA has rules about credits in the BOINC licensing you can award whatever you want. It stands to reason that higher credit will gain you more crunchers, which in turn allow more science to be completed at a faster rate.
me@rescam.org
ID: 9402 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profilemscharmack
Avatar

Send message
Joined: 4 Dec 07
Posts: 45
Credit: 1,257,904
RAC: 0
1 million credit badge14 year member badge
Message 9407 - Posted: 30 Jan 2009, 6:06:51 UTC
Last modified: 30 Jan 2009, 6:13:51 UTC

Yes, ABC@Home has the right thoughts to longer work units. They offer up to 1000 credits so people will crunch the longer ones, otherwise they will just about them. Good thought on their part. I've started back at SETI@home and GPU@home, however I'm using CUDA GPU's and they complete SETI work unit in about 30 seconds to 3 minutes and GPU@home in about 12 hours. Oh yeah, the points are good too.
ID: 9407 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
ProfilePaul D. Buck

Send message
Joined: 12 Apr 08
Posts: 621
Credit: 161,934,067
RAC: 0
100 million credit badge14 year member badge
Message 9411 - Posted: 30 Jan 2009, 7:24:01 UTC - in response to Message 9375.  

ABC also awards some 1K credit tasks. I don't know what the criteria is for selection, but I have gotten several where I asked for, say, 300 and got the 1K award... Not complaining, just commenting...

ABC rewards 1K credits (maximum!) to the longest WUs; their credit policy is explained here. ;-)


That is why I have always done at least some ABC ... CPDN I like because they are the only project that has 100+ hour tasks that a) uses trickles, and b) pays for partial completions because they know that models crash for many reasons ...

I stopped doing Sztaki because a) They have 100+ hour models, and b) you get paid nothing for a crash, and c) the project types don't acknowledge the problem

I also stopped doing some model types at Lattice for the same reason ... they run forever and abend or crash and you lose hundreds of hours of effort. and it is a known issue with the models. When models are short and stable the draconian no credit rule makes sense and can be accepted as a loss ... but ... sorry, if the models run too long and crash a lot I am going to find somewhere else to put my effort.

Prime Grid has had similar issues with some of their applications. ONe of the reasons I stopped doing some of their sub-projects and was putting major effort into Reisel Sieve till they disappeared ... though they may be back some day ...
ID: 9411 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
BarryAZ

Send message
Joined: 1 Sep 08
Posts: 520
Credit: 301,627,200
RAC: 12,174
300 million credit badge14 year member badgeextraordinary contributions badge
Message 9446 - Posted: 31 Jan 2009, 18:06:08 UTC - in response to Message 9349.  

But the awards for a system with a 'cuda video card using the SETI stock application are in some ways off the chart. So is DA going to act on that? Of course not -- and he shouldn't.

MilkyWay is seeing one of those upward spikes in credits for a couple of reasons, first, the actual user count has accelerated over the past month or so (430 joined in December, more than double that in January). Second, there seem to be more folks using optimized applications and those seriously push up the credit numbers.

I agree with the comments of others about higher credits for alpha/beta projects and for those who suggest that projects doing actual research and science should be comfortable with a higher credit structure. SETI is neither alpha/beta nor are they doing real science and research. I'm not saying the potential work done at SETI is without value, but I figure trotting out both the alpha/beta and science/research arguments should provide some decent 'talking points' in putting the DA 'SETI is the standard' arguments in their proper perspective.


However, remember DA wants Seti to be the standard. So how does MW compare to Seti stock, or how does Einstein stock compare to Seti stock? It is possible Einstein is a lower paying project than Seti.


ID: 9446 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
bobgoblin

Send message
Joined: 8 Dec 07
Posts: 60
Credit: 67,028,931
RAC: 0
50 million credit badge14 year member badge
Message 9516 - Posted: 1 Feb 2009, 12:10:24 UTC - in response to Message 9446.  

Second, there seem to be more folks using optimized applications and those seriously push up the credit numbers.



I very much agree with this comment. Currently I feel penalized for not running a recompiled app. I see machines much slower than my i7 completing units in 1/3rd the time getting 2 to 3 times the credit. is the credit cap still in effect?

Until we're all running the same applications, how can any adjustments to awarded credit be made?

Currently I've switched my i7 over to abc exclusively since, honestly, i get more credit there than with mw. but once there is a standard app for vista 64, i'll want to start mw on it again.
ID: 9516 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
BarryAZ

Send message
Joined: 1 Sep 08
Posts: 520
Credit: 301,627,200
RAC: 12,174
300 million credit badge14 year member badgeextraordinary contributions badge
Message 9534 - Posted: 1 Feb 2009, 21:35:40 UTC - in response to Message 9516.  

It seems to me that reducing the credit per completed work unit *because* individually optimized applications which are not readily available are too efficient is not the best approach.

Now as the work done by the optimization dudes (and dudettes) is either migrated into the stock client or is readily available to all, then credit changes can be justified.

I tend to compare the awards for the stock client here to the optimized non-Astropulse SETI client because that optimized client is readily obtained and readily deployed by nearly anyone. I realize the 'stock application' over in SETI land (when they are running, they have had a rather 'beta-ish' run of reliability problems over the past couple of weeks) is less efficient, but the optimized client is quite easy to get and quite easy to deploy -- so anyone can use it instead.

As to those doing optimization of the Milkyway client -- I figure any who mourn the frequent changes to the released application making it more difficult to do what they are doing -- well, this is a alpha status project for one, and for two, Travis seems quite willing to implement the solid work of a number of the optimization folks in order to 'share the wealth' -- as that increases the amount of actual science that gets done. So, JUST LIKE THE OPTIMIZED APPLICATIONS, the released client is very much something of a moving target. That should neither be a surprise nor be viewed as something to whinge about.
ID: 9534 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Misfit
Avatar

Send message
Joined: 27 Aug 07
Posts: 915
Credit: 1,503,319
RAC: 0
1 million credit badge15 year member badge
Message 9587 - Posted: 3 Feb 2009, 1:36:38 UTC

Running MW and Cosmo equally I see Cosmo has taken the lead in RAC. :/
me@rescam.org
ID: 9587 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
ProfileArion
Avatar

Send message
Joined: 10 Aug 08
Posts: 218
Credit: 41,846,854
RAC: 0
30 million credit badge14 year member badge
Message 9602 - Posted: 3 Feb 2009, 8:59:12 UTC - in response to Message 9349.  

Arion.

Based on your numbers:

MW - 31.72 credits / hour.
EH - 15.50 credits / hour.

However, remember DA wants Seti to be the standard. So how does MW compare to Seti stock, or how does Einstein stock compare to Seti stock? It is possible Einstein is a lower paying project than Seti.


Sorry for taking so long to respond, but my main system decided a couple of weeks ago to take a powder and its been a nightmare ever since. Was in the middle of End of Year reports for a company I contract out to and I've been going nuts trying to trouble shoot and replace the system. Finally got the accounting and reports done Sunday night and out in today's mail. (new MB now as well)

I wasn't sure if I was doing the math right or not. Switching between websites looking at results and trying to make sure I was really giving the correct figures and a balanced accounting of them (as per Travis' reason for this thread) and reading what I was scribbling on a scrap piece of paper just added to my confusion. <smile>

I've just given up with all the credit crapola. I started off picking projects for what I perceived as scientific in value that I was personally interested in. Did it that way from day 1 back in Aug 1999 when I joined the seti@home project. In almost 9 years of this I've ONLY worked with 4 projects. I'd say it was about a year or so ago I really felt like my contributions were worthless compared to what others using faster computers, super farms, all manners of optimized clients (approved and not) were doing when I watched what 1000, 2000, 3000 credit days kept getting shoved lower and lower down the stats list. When milksap released his app in Oct I figured why not try to make up some of the losses over the past few years and jumped on the bandwagon. Little did I know what I was getting into by going the credit is king route!!!! There's the usual arguements about fairness in intra projects and inter projects and the fight goes on. At the rate it is going (based on credits) I can see me running the same amount of time processing as I was 9 years ago and getting NO credits at the rate things with all the projects that I am currently running. So it comes down to 1) keep doing it for less credits 2) save myself a chunk of change and quit all together like 95% of the people in my sign up day or 3) forget credits and do it because I enjoy it. I really haven't decided where I want to go. But I know my wife would really prefer that I picked #2!! <smile>

How does that saying go? And the beat goes on....



ID: 9602 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
ProfileGavin Shaw
Avatar

Send message
Joined: 16 Jan 08
Posts: 98
Credit: 1,371,299
RAC: 0
1 million credit badge14 year member badge
Message 9603 - Posted: 3 Feb 2009, 9:42:48 UTC

Arion.

No problems.

I also have a tendency to pick projects that I deem to be worthwhile in my view. I started with Seti but have moved away for reasons. I do Rosetta because I believe it has value. I do MW because it aligns with my interest in space/astronomy. I have one machine doing some SIMAP because I see that SIMAP has value.

At least my wife lets me run Boinc without complaint. SHe even lets me run it on her laptop and will let me run it on the system the uni will assign her when she starts her PhD.

Never surrender and never give up. In the darkest hour there is always hope.

ID: 9603 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
David @ TPS

Send message
Joined: 20 Nov 08
Posts: 24
Credit: 2,561,361
RAC: 0
2 million credit badge13 year member badge
Message 9607 - Posted: 3 Feb 2009, 14:16:32 UTC

For what it is worth:

P-4 HT, 3.2 Ghz, XP Pro.

SETI - 204 seconds per credit (optimized app)
MW - 312 seconds per credit
E@H - 388 seconds per credit

(above rounded to nearest whole number)

These figures are for a single thread of the HT.

Dave
ID: 9607 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
John Clark

Send message
Joined: 4 Oct 08
Posts: 1734
Credit: 64,228,409
RAC: 0
50 million credit badge13 year member badge
Message 9608 - Posted: 3 Feb 2009, 14:24:54 UTC

Perhaps this thread should take note of Cluster Physik's comment here on flop counting as a contribution towards the credit awarded by a project and any comparisons between project credit per hour (or what ever).

This indicates the flop count work dome in this project is more than the base line SETI project. Could that be a justification for any higher credit award here as we are doing more work, not like-for-like?
ID: 9608 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profilebanditwolf
Avatar

Send message
Joined: 12 Nov 07
Posts: 2425
Credit: 524,164
RAC: 0
500 thousand credit badge14 year member badge
Message 9609 - Posted: 3 Feb 2009, 14:53:54 UTC - in response to Message 9608.  

Perhaps this thread should take note of Cluster Physik's comment here on flop counting as a contribution towards the credit awarded by a project and any comparisons between project credit per hour (or what ever).

This indicates the flop count work dome in this project is more than the base line SETI project. Could that be a justification for any higher credit award here as we are doing more work, not like-for-like?


Cluster's comment sounds right. Twice the precision should equal more credits, not the ever ending less credits. I think DA needs to learn what he is talking about before getting on other projects.
Doesn't expecting the unexpected make the unexpected the expected?
If it makes sense, DON'T do it.
ID: 9609 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Cluster Physik

Send message
Joined: 26 Jul 08
Posts: 627
Credit: 94,940,203
RAC: 0
50 million credit badge14 year member badgeextraordinary contributions badge
Message 9610 - Posted: 3 Feb 2009, 15:14:46 UTC - in response to Message 9609.  
Last modified: 3 Feb 2009, 15:16:47 UTC

Perhaps this thread should take note of Cluster Physik's comment here on flop counting as a contribution towards the credit awarded by a project and any comparisons between project credit per hour (or what ever).

This indicates the flop count work dome in this project is more than the base line SETI project. Could that be a justification for any higher credit award here as we are doing more work, not like-for-like?


Cluster's comment sounds right. Twice the precision should equal more credits, not the ever ending less credits. I think DA needs to learn what he is talking about before getting on other projects.

Maybe I should have added that the current credit multiplier at SETI is 2.77 credits/TFlop. So even when doubling it for double precision, we are looking at ~7 credits for the single stream WUs and ~10.5 credits for the dual stream WUs.

That would not be a problem for the optimized apps, because Travis could finally take down the stupid creditlimit and one would still get roughly the same or even a bit more on fast CPUs. If the project starts to use a better compiler, it is also not a problem for the stock app (just look at the linux versions compiled by speedimic, no code changes involved but massive speedup), the credit level should be approximately the same.

Furthermore this is totally necessary for the GPU app (that hopefully will come) and enables an easy and somewhat fair base of comparison with other projects. I guess even DA cannot complain if one is using flops counting (he suggested that method quite often) and MW is just more efficient than the SETI stock application.
ID: 9610 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profilebanditwolf
Avatar

Send message
Joined: 12 Nov 07
Posts: 2425
Credit: 524,164
RAC: 0
500 thousand credit badge14 year member badge
Message 9613 - Posted: 3 Feb 2009, 15:49:00 UTC - in response to Message 9610.  

What does this make the wu's worth. More/less than they are? I'm thinking more right?
Doesn't expecting the unexpected make the unexpected the expected?
If it makes sense, DON'T do it.
ID: 9613 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
ProfileKalessin
Avatar

Send message
Joined: 10 Nov 07
Posts: 42
Credit: 27,012,695
RAC: 0
20 million credit badge14 year member badge
Message 9628 - Posted: 3 Feb 2009, 20:29:29 UTC - in response to Message 9610.  
Last modified: 3 Feb 2009, 20:37:36 UTC

Maybe I should have added that the current credit multiplier at SETI is 2.77 credits/TFlop. So even when doubling it for double precision, we are looking at ~7 credits for the single stream WUs and ~10.5 credits for the dual stream WUs.


What does this make the wu's worth. More/less than they are? I'm thinking more right?


Since they are at a little above 10 and around 16 on my fast Quad, it would be definetly less!
And that would be what would make MW be comparable to other projects.
And furthermore the creditlimit would really need to be removed. It has truly become obsolete very soon after it was introduced, since there were boinc versions out there that simulated being so many PCs that their output was never reduced.
Dragons can fly because they don't fit into pirate ships!
ID: 9628 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Previous · 1 . . . 11 · 12 · 13 · 14 · 15 · Next

Message boards : Number crunching : credit comparison to other projects

©2022 Astroinformatics Group