Observation of CreditNew Impact

Message boards : Number crunching : Observation of CreditNew Impact
Message board moderation

To post messages, you must log in.

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

AuthorMessage
Keith White
Avatar

Send message
Joined: 29 May 99
Posts: 392
Credit: 13,035,233
RAC: 22
United States
Message 1384151 - Posted: 24 Jun 2013, 11:14:53 UTC - in response to Message 1384143.  

My RAC partially recovered due to me doing AP WUs for the first time.
"Life is just nature's way of keeping meat fresh." - The Doctor
ID: 1384151 · Report as offensive
ExchangeMan
Volunteer tester

Send message
Joined: 9 Jan 00
Posts: 115
Credit: 157,719,104
RAC: 0
United States
Message 1384165 - Posted: 24 Jun 2013, 12:04:10 UTC - in response to Message 1384151.  

My RAC partially recovered due to me doing AP WUs for the first time.

Same thing happened here also. Not sure how much is attributable to AP or we all hit bottom and there's nowhere else to go.

ID: 1384165 · Report as offensive
Profile bj

Send message
Joined: 11 Oct 00
Posts: 163
Credit: 50,429,507
RAC: 0
United States
Message 1384167 - Posted: 24 Jun 2013, 12:34:13 UTC

RAC went from 80k to 52k and still dropping.

Where is that bottom that they all is talking about?

bj
ID: 1384167 · Report as offensive
Profile HAL9000
Volunteer tester
Avatar

Send message
Joined: 11 Sep 99
Posts: 6534
Credit: 196,805,888
RAC: 57
United States
Message 1384176 - Posted: 24 Jun 2013, 13:22:25 UTC - in response to Message 1384167.  

RAC went from 80k to 52k and still dropping.

Where is that bottom that they all is talking about?

bj

I'm not exactly sure. I figure that my CPU MB only machines are just going for a visit to the Titanic.
However the dots on the graphs are getting closer. So instead of dropping 7-8% a day it is now only about 2-3%.
SETI@home classic workunits: 93,865 CPU time: 863,447 hours
Join the [url=http://tinyurl.com/8y46zvu]BP6/VP6 User Group[
ID: 1384176 · Report as offensive
Cruncher-American Crowdfunding Project Donor*Special Project $75 donorSpecial Project $250 donor

Send message
Joined: 25 Mar 02
Posts: 1513
Credit: 370,893,186
RAC: 340
United States
Message 1384179 - Posted: 24 Jun 2013, 13:42:42 UTC - in response to Message 1384165.  

My RAC partially recovered due to me doing AP WUs for the first time.

Same thing happened here also. Not sure how much is attributable to AP or we all hit bottom and there's nowhere else to go.


It's ALL AP - they give much more credit per unit time of execution.

Note that there are very few APs being given out now, too, which doesn't help RAC at all.
ID: 1384179 · Report as offensive
Terror Australis
Volunteer tester

Send message
Joined: 14 Feb 04
Posts: 1817
Credit: 262,693,308
RAC: 44
Australia
Message 1384282 - Posted: 24 Jun 2013, 17:20:11 UTC - in response to Message 1384179.  

It's ALL AP - they give much more credit per unit time of execution.

Note that there are very few APs being given out now, too, which doesn't help RAC at all.

It's not all AP. The computer I mentioned above is MB only, and the credits per WU are definitely increasing. According to FreeDC the credit per day rating of this box has risen by 5k since last week.

Even V7 VHARS, which take twice as long to run as a V6 are now paying close to twice the credit. There are still some anomalies but things are definitely on the improve.

T.A.
ID: 1384282 · Report as offensive
Keith White
Avatar

Send message
Joined: 29 May 99
Posts: 392
Credit: 13,035,233
RAC: 22
United States
Message 1384356 - Posted: 24 Jun 2013, 20:31:17 UTC - in response to Message 1384165.  

My RAC partially recovered due to me doing AP WUs for the first time.

Same thing happened here also. Not sure how much is attributable to AP or we all hit bottom and there's nowhere else to go.


Well AP CPU 32 credits/hr while V7 CPU is only 20. AP GPU is 175 credits/hr while V7 GPU is only 50.

Back in the day MB V6 CPU was 35 credits/hr wand MB V6 GPU was 110-120 credits/hr.

Without AP I would still be going down to a RAC around 2000.

"Life is just nature's way of keeping meat fresh." - The Doctor
ID: 1384356 · Report as offensive
Lionel

Send message
Joined: 25 Mar 00
Posts: 680
Credit: 563,640,304
RAC: 597
Australia
Message 1384418 - Posted: 25 Jun 2013, 1:44:16 UTC - in response to Message 1384282.  

It's ALL AP - they give much more credit per unit time of execution.

Note that there are very few APs being given out now, too, which doesn't help RAC at all.

It's not all AP. The computer I mentioned above is MB only, and the credits per WU are definitely increasing. According to FreeDC the credit per day rating of this box has risen by 5k since last week.

Even V7 VHARS, which take twice as long to run as a V6 are now paying close to twice the credit. There are still some anomalies but things are definitely on the improve.

T.A.


That may be the case but at my end I have not seen any increase in credits granted. When I looked yesterday, credit being granted for a v7 WU was still dismally lower than for a v6 WUs. It is true that if you add AP into your processing mix, then RAC will increase which is why I have stopped posting my daily totals as I have now added AP into my mix and this will cloud the actual impact of v7 credits.

All up,
<colourful language>
they could not have stuffed it up more, even if they had tried too.
</colourful language>

The lack of acknowledgement and the silence around the impact (from its creators and the project owner) indicates to me that they have no idea. The solution, I believe, that will be adopted will be constant ongoing tinkering until they think it is about right. So expect intermittent, small incremental gains over time to daily totals, etc.

I also doubt that any solution will be retro fitted to v7 WUs processed thus far, even tough in theory it should be.

cheers




ID: 1384418 · Report as offensive
Profile Bernie Vine
Volunteer moderator
Volunteer tester
Avatar

Send message
Joined: 26 May 99
Posts: 9958
Credit: 103,452,613
RAC: 328
United Kingdom
Message 1384545 - Posted: 25 Jun 2013, 13:08:16 UTC

The lack of acknowledgement and the silence around the impact (from its creators and the project owner) indicates to me that they have no idea.


So you do not consider these post by Dr Eric Korpela's to be any sort of acknowledgement

http://setiathome.berkeley.edu/forum_thread.php?id=71805&postid=1378218

The fall in credit was most definitely not deliberate. I wouldn't even know how to cause such a thing using the BOINC credit new. I had assumed it was a case of "fast machines report results first" which usually drives the granted credit to low values when a new version is released, but there should have been ample time for that to fix itself.

Of course the corollary is that I haven't figured out how to fix it yet.

My RAC has gone down by less than 10%. Is this worse for people using anonymous platform?


http://setiathome.berkeley.edu/forum_thread.php?id=71889&postid=1381896

I suppose I should make a couple points.

First, "CreditNew" has been implemented at SETI@home for a year now, so this reduction in credit is not due to implementing "CreditNew" since it was implemented ages ago.

Second, how credits are normalized hasn't changed between v6 and v7 under CreditNew. Since our results are mostly stock Windows and CUDA under Windows, those results set the normalization.

Based upon the archives, the S@H 6 windows_intel app was generating an average of 0.00556 credits per elapsed second. The current S@H 7 windows_intel app is generating 0.00514 credits per elapsed second, or about 7% less. Astropulse 6 is currently generating 0.00491 credits per elapsed second or about 12% less than S@H v6. I'm hoping the Astropulse issue is resolving itself. (It appears to be slowly coming back to normal.)

There aren't a whole lot of knobs I can turn to get that 7% back. I've bumped the estimated GPU efficiency by 20% in hopes that that would help, but thus far I haven't seen a change. I'm going to try increasing the workunit work estimates slowly, but I think that change will get normalized out.

Yes, there are projects that offer more credit. A number of projects have chosen to detach their credits from any measure of actual processing, usually by pretending they are getting 100% efficiency out of GPUs. I fought that battle for years, and lost. Those projects have entirely devalued the BOINC credit system.



ID: 1384545 · Report as offensive
Lionel

Send message
Joined: 25 Mar 00
Posts: 680
Credit: 563,640,304
RAC: 597
Australia
Message 1384669 - Posted: 25 Jun 2013, 23:24:17 UTC - in response to Message 1383498.  
Last modified: 25 Jun 2013, 23:24:44 UTC

Bernie,

I note that there are 3 things in the first sentence of which you focus your first comment on only one of these. I would have to say that an abstract comment basically saying "oops" on 7 Jun does not constitute acknowledgement of the problem. I do note that there is nothing around "I apologise for...", "I am sorry for..." in the comment to the contributor base, nor has there been any further real communication around the issue, what they are doing and how it is progressing (a note on 16 Jun, 9 days after the first comment, does not constitute communication). I also note the comment "I haven't figured out how to fix it yet" which agrees with the last thought of the first sentence, that is, "that they have no idea".

I also draw your attention to his comment, "My RAC has gone down by less than 10%. Is this worse for people using anonymous platform?". In case you haven’t noticed, I have been posting my daily totals in this thread. I posted my daily totals from 16 May to 31 May when I was running on v6. The daily average over the period was 221,878. On 1 Jun, I migrated all my boxes to v7 and aborted all waiting v6 work. From 1 Jun to 22 Jun I recorded the daily totals (you can see them in a thread above). The average was 102,376. This is 46.1% of my previous running average when only doing v6. Many others have had similar experiences with falling daily totals with v7.

In short, the credit system is broken for v7 MB WUs. This is my view and I appreciate and accept that you and/or others may not agree fully or otherwise.

I do note that you talk about RAC not falling that much. RAC is a lagging indicator. If you want to see the actual impact to yourself you should go to either BOINC Stats or Free-DC and chart your credits over time.

The comments enclosed within the second side barred text are not fully correct. If they were, then the numbers that I was posting in this thread would not be as low as they are. There is an anomaly between what he is looking at and what people are experiencing. The investigation of the anomaly should lead them to the answer.

In relation to the last comment around other projects offering more credit, it has no impact on me as I do not do work for other projects. I am like others who only crunch Seti.


To Dr Eric Korpela

I offer the following (with all emotion aside):

Your comment: “Second, how credits are normalized hasn't changed between v6 and v7 under CreditNew. Since our results are mostly stock Windows and CUDA under Windows, those results set the normalization.”

Firstly I would ask you to look at the daily totals posted by me in this thread. Given that all machines moved to v7 on 1 June, you can see that daily totals reduced to around 46% of previous daily totals under v7 compared to v6 (54% reduction).

The machines in question all ran optimised apps for Multi-beam only over the period 1 June to 22 June.


Your comment: “Based upon the archives, the S@H 6 windowsintel app was generating an average of 0.00556 credits per elapsed second. The current S@H 7 windowsintel app is generating 0.00514 credits per elapsed second, or about 7% less. Astropulse 6 is currently generating 0.00491 credits per elapsed second or about 12% less than S@H v6. I'm hoping the Astropulse issue is resolving itself. (It appears to be slowly coming back to normal.)”

It would appear that the numbers are based on averages from the database. Can I suggest that rather than looking at database averages which can have a tendency to mask issues, that investigation look at machines and work units that have run on optimised and non-optimised applications. I would also offer the following observation to you. I have ceased processing v7 MB work units on 2 of the machines (the third machine will cease processing v7 MB in a day or two). Since doing this, daily totals have risen sharply and significantly on both of these computers. These machines are currently only working on AP across CPU and GPU.

Please feel free to look at the machines that I have, and the credits granted against work units if this helps.

If you wish to contact me to discuss any of this I would more than welcome your approach to help resolve the anomaly as it can only benefit all.

rgds

Lionel
ID: 1384669 · Report as offensive
Grant (SSSF)
Volunteer tester

Send message
Joined: 19 Aug 99
Posts: 13855
Credit: 208,696,464
RAC: 304
Australia
Message 1384730 - Posted: 26 Jun 2013, 5:58:15 UTC - in response to Message 1384669.  


My optimised v7 MB only systems RACs continue to fall.
Running out of work overnight probably hasn't helped things.
Grant
Darwin NT
ID: 1384730 · Report as offensive
Sleepy
Volunteer tester
Avatar

Send message
Joined: 21 May 99
Posts: 219
Credit: 98,947,784
RAC: 28,360
Italy
Message 1384760 - Posted: 26 Jun 2013, 8:35:52 UTC - in response to Message 1384669.  


I also draw your attention to his comment, "My RAC has gone down by less than 10%. Is this worse for people using anonymous platform?".


With immense respect to all SETI developers and in particular to the people working at the labs, I would add to your comment (with whom I agree regarding data and perspective. I walked the same path) that, as I already pointed out some posts above, Eric was comparing the throughput of stock applications.

This means he was basically comparing V6 stock application vs. V7 stock, which is however heavily optimised, even if not fully.
To put things in an easier way, though partially incorrect, it is a bit like comparing V6 stock against V6 optimised.
If stock and optimised give the same credits, it means that now credits given for same amount of work done are cut in half, since optimised are roughly double (or more) efficient respect to stock.
This is the reason why Eric sees things almost staying the same.

If he basically switched from full paying/half efficient stock V6 to half paying/full efficient optimised (though stock) V7, then it is quite reasonable that the balance is roughly even (though, even so, there is a 10% reduction).
Who was and is on optimised, of course had nothing to gain switching, since he/she already was using the best software and thus found his/her credit cut in half.

Of course this has nothing to do with the scientific work done. The project as a whole is now more efficient and does more accurate analysis.
Though many people will tend to switch to AP and maybe processing power of MB will be reduced (this should be an "antropological" migration that should be observed form the cockpit, especially if the big crunchers will go this way, if not already there).

Happy crunching!

Sleepy
ID: 1384760 · Report as offensive
Profile Warren Kozey
Avatar

Send message
Joined: 6 Jul 99
Posts: 54
Credit: 5,026,721
RAC: 0
Canada
Message 1384885 - Posted: 26 Jun 2013, 17:49:20 UTC - in response to Message 1384760.  

RAC score is totally screwed since updating to seti v7 I am running 5 machines and was averaging about 5000 RAC per day running seti v6.x since upgrading to v7 I have dropped to about 1900 per day and still dropping was running lunatics v.40 and now v.41
Gimme BEER and WU's!!!!
ID: 1384885 · Report as offensive
Lionel

Send message
Joined: 25 Mar 00
Posts: 680
Credit: 563,640,304
RAC: 597
Australia
Message 1385019 - Posted: 27 Jun 2013, 4:09:07 UTC - in response to Message 1384885.  


You are not alone here, every one is suffering. In my case, I saw a 54% decline in daily credit after moving to v7.


ID: 1385019 · Report as offensive
Lionel

Send message
Joined: 25 Mar 00
Posts: 680
Credit: 563,640,304
RAC: 597
Australia
Message 1385095 - Posted: 27 Jun 2013, 12:52:11 UTC - in response to Message 1385092.  



I agree and do not think that you are wrong.




ID: 1385095 · Report as offensive
kittyman Crowdfunding Project Donor*Special Project $75 donorSpecial Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 9 Jul 00
Posts: 51478
Credit: 1,018,363,574
RAC: 1,004
United States
Message 1385127 - Posted: 27 Jun 2013, 15:58:16 UTC

It is what it is, what it is, what it is.

I am done fretting about my RAC, and at this point just consider it the new norm.
It shall probably do some further self adjusting as time goes on.

It shall not affect the amount of work I do for the project, and that is the most important thing.

RAC for me is still relevant as a benchmark against all other participants on the Seti project. And I don't care how that relates to the credits that other projects wish to hand out like candy.

Meow.
"Time is simply the mechanism that keeps everything from happening all at once."

ID: 1385127 · Report as offensive
Profile Donald L. Johnson
Avatar

Send message
Joined: 5 Aug 02
Posts: 8240
Credit: 14,654,533
RAC: 20
United States
Message 1385142 - Posted: 27 Jun 2013, 16:50:22 UTC - in response to Message 1385092.  


You are not alone here, every one is suffering. In my case, I saw a 54% decline in daily credit after moving to v7.

My concern isn't the drop in the RAC (we are all in the same boat & perhaps this will be the New Normal). But...

A lack of real response from those managing this Project.

There HAVE been real responses from Dr. K and the v7 developers, in THIS THREAD and others. You apparently have not seen them.

Here's a summary of what you have missed:

CreditNew is a function of BOINC, not Seti@Home. Eric has said he has neither the knowledge not the authority to change CreditNew.

Because the v7 science apps do more analysis, they were expected to take longer to process, so tasks per day and RAC were expected to drop, but the change in credit per task was NOT expected to be as large as it seems to have been. This HAS been brought to Dr. Anderson's attention, but we really don't expect him to make any changes.

Eric and others have said it may take as long as 5-6 weeks after roll-out for things to settle out and a new "normal" establish itself. We are not there yet, but reports are that credit grants are beginning to return to v6 levels.

So for at least the next couple of weeks, the old rule still applies:
Here at Seti@Home, patience is not just a virtue, it is a REQUIREMENT.

Donald
Infernal Optimist / Submariner, retired
ID: 1385142 · Report as offensive
Grant (SSSF)
Volunteer tester

Send message
Joined: 19 Aug 99
Posts: 13855
Credit: 208,696,464
RAC: 304
Australia
Message 1385171 - Posted: 27 Jun 2013, 18:07:44 UTC - in response to Message 1385142.  


On my slowest cruncher (and the one that had a couple of days downtime & problems getting work after timing out WUs) RAC did rebound slightly, but has started to fall aagin, but is not quite back down to it's lowest point so far. On my faster system, RAC continues to fall.
Grant
Darwin NT
ID: 1385171 · Report as offensive
Lionel

Send message
Joined: 25 Mar 00
Posts: 680
Credit: 563,640,304
RAC: 597
Australia
Message 1385249 - Posted: 27 Jun 2013, 23:16:20 UTC - in response to Message 1385127.  

It is what it is, what it is, what it is.

I am done fretting about my RAC, and at this point just consider it the new norm.
It shall probably do some further self adjusting as time goes on.

It shall not affect the amount of work I do for the project, and that is the most important thing.

RAC for me is still relevant as a benchmark against all other participants on the Seti project. And I don't care how that relates to the credits that other projects wish to hand out like candy.

Meow.




Mark, firstly let me say that I appreciate your view and position.

I (and undoubtedly others) consider there to be a problem. Where that problem is and how it manifests itself is beyond my ability (and that of many others) to analyse and determine, but not beyond the ability of those that created the system. The issue as to whether they have the will or desire to do this and understand the issue is another question.

I, like you, only crunch Seti, so cross project credit rates are of no issue or interest to me either.

The issue of benchmarking within the project is, as I agree with you, relevant here. To that end, I started conducting my own “simple test” a few days ago. The outcome of this may be of interest to some, but as it looks at the moment, it would appear to show that there is an issue in here as well.

Lionel


ID: 1385249 · Report as offensive
Profile doublechaz

Send message
Joined: 17 Nov 00
Posts: 90
Credit: 76,455,865
RAC: 735
United States
Message 1385282 - Posted: 28 Jun 2013, 4:01:53 UTC

Seems to me the credit has varied a good deal with each new version of the app going back more than ten years. Who cares? We're all in this project, the only place where scores for this project have any meaning.

I came here to see if I could find out why my RAC is half what it used to be. The answer is that everyone is seeing the same. It's an artifact of the version change. Nothing broken. Ok. Question answered.

Nothing to whine about. Thanks everyone for clearing that up.


ID: 1385282 · Report as offensive
Previous · 1 . . . 11 · 12 · 13 · 14 · 15 · Next

Message boards : Number crunching : Observation of CreditNew Impact


 
©2024 University of California
 
SETI@home and Astropulse are funded by grants from the National Science Foundation, NASA, and donations from SETI@home volunteers. AstroPulse is funded in part by the NSF through grant AST-0307956.