Welcome to MilkyWay@home

New MilkyWay Separation Modified Fit Runs

Message boards : News : New MilkyWay Separation Modified Fit Runs
Message board moderation

To post messages, you must log in.

Previous · 1 · 2

AuthorMessage
Profile mikey
Avatar

Send message
Joined: 8 May 09
Posts: 3339
Credit: 524,010,781
RAC: 0
Message 60449 - Posted: 21 Nov 2013, 14:44:55 UTC - in response to Message 60279.  

Thanks for bringing it up, it is curious that modified fit without the new background takes much longer than regular separation. I will look into it.

Jake W.


For some reason my credits HAVE been going up the last few days, if you 'fixed' something THANK YOU!! If not then something is even stranger!!
ID: 60449 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile mikey
Avatar

Send message
Joined: 8 May 09
Posts: 3339
Credit: 524,010,781
RAC: 0
Message 60460 - Posted: 25 Nov 2013, 15:40:20 UTC - in response to Message 60449.  

Thanks for bringing it up, it is curious that modified fit without the new background takes much longer than regular separation. I will look into it.

Jake W.


For some reason my credits HAVE been going up the last few days, if you 'fixed' something THANK YOU!! If not then something is even stranger!!


I spoke too soon, the credits are now headed back down again!!! All of my pc's rac have gone down, up and are now headed back down again!! In the space of one month one pc went from 9k to 8.4k to 9.4k and now back down to 8.9k per day!! Nothing has changed except the units, the pc has NOT been turned off even once and runs 24/7 full time crunching units! It's alot like you guys are fighting with the 'credit new' system and haven't figured it out yet.
ID: 60460 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Link
Avatar

Send message
Joined: 19 Jul 10
Posts: 627
Credit: 19,321,489
RAC: 1,827
Message 60471 - Posted: 28 Nov 2013, 10:07:55 UTC - in response to Message 60460.  

I spoke too soon, the credits are now headed back down again!!! All of my pc's rac have gone down, up and are now headed back down again!! In the space of one month one pc went from 9k to 8.4k to 9.4k and now back down to 8.9k per day!! Nothing has changed except the units, the pc has NOT been turned off even once and runs 24/7 full time crunching units! It's alot like you guys are fighting with the 'credit new' system and haven't figured it out yet.

Modified Fit does not seem to use CreditNew, if that was the case not all WUs from the same batch would have exactly the same credit, for example 213.76. There's a good reson, why CreditNew is also called "random credit generator".

And do you really want to complain about the RAC going from 9.0k to 8.9k?
ID: 60471 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile mikey
Avatar

Send message
Joined: 8 May 09
Posts: 3339
Credit: 524,010,781
RAC: 0
Message 60473 - Posted: 28 Nov 2013, 12:20:27 UTC - in response to Message 60471.  

I spoke too soon, the credits are now headed back down again!!! All of my pc's rac have gone down, up and are now headed back down again!! In the space of one month one pc went from 9k to 8.4k to 9.4k and now back down to 8.9k per day!! Nothing has changed except the units, the pc has NOT been turned off even once and runs 24/7 full time crunching units! It's alot like you guys are fighting with the 'credit new' system and haven't figured it out yet.

Modified Fit does not seem to use CreditNew, if that was the case not all WUs from the same batch would have exactly the same credit, for example 213.76. There's a good reson, why CreditNew is also called "random credit generator".

And do you really want to complain about the RAC going from 9.0k to 8.9k?


No, but it is now down to 8.4 and on a steep declining angle. I know rac goes up and down and depends on a bazillion things, most way beyond my control. The point was that with all the new runs perhaps they had forgotten to account for the credit changes they had made in the last run to make them come up.

Thanks for explaining the 'creditnew' randomness. I did not know that, but thinking about it, after your explanation, should have! Hmmm...I am getting old!!
ID: 60473 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Link
Avatar

Send message
Joined: 19 Jul 10
Posts: 627
Credit: 19,321,489
RAC: 1,827
Message 60477 - Posted: 28 Nov 2013, 18:24:12 UTC - in response to Message 60473.  
Last modified: 28 Nov 2013, 18:38:28 UTC

No, but it is now down to 8.4 and on a steep declining angle. I know rac goes up and down and depends on a bazillion things, most way beyond my control. The point was that with all the new runs perhaps they had forgotten to account for the credit changes they had made in the last run to make them come up.

I always look at credit/month, that's pretty much the only usable value or I calculate the RAC myself based on some WUs, that's another way to figure out, if some WUs pay less (that has never happen so far for me and my HD3850 here at Milkyway). Any other values I'd not trust, specially not the RAC.

As a little additional thing to what you posted before, i.e. 9k to 8.4k to 9.4k and than back to ~9k: that's pretty much what I'd expect to see when new runs are started, because each time that happens, we have higher rate of inconclusive results (at least for the "normal" separation, but I assume it's not much different for the modified one) and hence we see these 3 steps:

1. At first the RAC goes down, because instead of validating instantly like usual, many task become inconclusive and need one or two additional results before validation -> less tasks validated, lower RAC.

2. After a while the inconclusive rate gets back to normal and the normal amount of tasks are validating instantly, but at the same time the additional inconclusive tasks from step 1 are validating -> validation rate above the normal level, RAC higher than normal.

3. Once everything settles down, you get back to the RAC you had before, or at least something very close to that. Note here, that it has been reported several times here in forum, that not all WUs are able to generate the same load on GPUs, specially on the higer-end ones, so sometimes the GPU is loaded little less and doing little less job per hour, specially if running just one WU at a time. That might result in slightly different RAC for each batch of WUs.

Step 3 might of course not occur if new runs are released often, there's a new thread about new modified runs.


Thanks for explaining the 'creditnew' randomness. I did not know that, but thinking
about it, after your explanation, should have! Hmmm...I am getting old!!

You might want to run some SETI WUs, I'm sure that will give you a good impression of what randomness we are talking here about. :D
ID: 60477 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile mikey
Avatar

Send message
Joined: 8 May 09
Posts: 3339
Credit: 524,010,781
RAC: 0
Message 60482 - Posted: 29 Nov 2013, 12:58:33 UTC - in response to Message 60477.  

No, but it is now down to 8.4 and on a steep declining angle. I know rac goes up and down and depends on a bazillion things, most way beyond my control. The point was that with all the new runs perhaps they had forgotten to account for the credit changes they had made in the last run to make them come up.

I always look at credit/month, that's pretty much the only usable value or I calculate the RAC myself based on some WUs, that's another way to figure out, if some WUs pay less (that has never happen so far for me and my HD3850 here at Milkyway). Any other values I'd not trust, specially not the RAC.

As a little additional thing to what you posted before, i.e. 9k to 8.4k to 9.4k and than back to ~9k: that's pretty much what I'd expect to see when new runs are started, because each time that happens, we have higher rate of inconclusive results (at least for the "normal" separation, but I assume it's not much different for the modified one) and hence we see these 3 steps:

1. At first the RAC goes down, because instead of validating instantly like usual, many task become inconclusive and need one or two additional results before validation -> less tasks validated, lower RAC.

2. After a while the inconclusive rate gets back to normal and the normal amount of tasks are validating instantly, but at the same time the additional inconclusive tasks from step 1 are validating -> validation rate above the normal level, RAC higher than normal.

3. Once everything settles down, you get back to the RAC you had before, or at least something very close to that. Note here, that it has been reported several times here in forum, that not all WUs are able to generate the same load on GPUs, specially on the higer-end ones, so sometimes the GPU is loaded little less and doing little less job per hour, specially if running just one WU at a time. That might result in slightly different RAC for each batch of WUs.

Step 3 might of course not occur if new runs are released often, there's a new thread about new modified runs.


Thanks for explaining the 'creditnew' randomness. I did not know that, but thinking
about it, after your explanation, should have! Hmmm...I am getting old!!

You might want to run some SETI WUs, I'm sure that will give you a good impression of what randomness we are talking here about. :D


I ONLY run the cpu units here, not the gpu units, but I think your premise still applies. As for Seti units I will not go back there for personal reasons, so that is out of the question for me. I do like the premise of Seti, but in March 2007 made the decision to leave and never look back. Too many other fish in the sea etc.
ID: 60482 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile valterc

Send message
Joined: 28 Aug 09
Posts: 23
Credit: 1,294,336,859
RAC: 219,449
Message 60522 - Posted: 5 Dec 2013, 14:43:51 UTC - in response to Message 60482.  

I just checked the credit issue again and found the same behavior as a few weeks ago. Credit is fixed at 213.76 while runtimes vary considerably (on my HD7970) from ~55s (ps_modfit_15_3s_128wrap_3_1382698503_6716313_2) to ~85s (de_modfit_15_3s_bplmodfit_128wrap_1_1382698503_6587065_3).

ID: 60522 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Previous · 1 · 2

Message boards : News : New MilkyWay Separation Modified Fit Runs

©2024 Astroinformatics Group