Correction Factor of 3.8286???

Message boards : Number crunching : Correction Factor of 3.8286???
Message board moderation

To post messages, you must log in.

AuthorMessage
CodeRedDewd

Send message
Joined: 22 Nov 09
Posts: 5
Credit: 2,854
RAC: 0
United States
Message 949137 - Posted: 22 Nov 2009, 22:44:51 UTC

It says that my Duration Correction Factor is 3.8286... What does that mean? From my understanding it means that my computer is much slower than what it predicts?

I have a Phenom 9950 @ 3.25Ghz 4gig ram and an XFX9800 GX2 overclocked as well. I thought my correction factor would be less than 1 for sure.
ID: 949137 · Report as offensive
Profile 52 Aces
Avatar

Send message
Joined: 7 Jan 02
Posts: 497
Credit: 14,261,068
RAC: 67
United States
Message 949140 - Posted: 22 Nov 2009, 22:59:29 UTC - in response to Message 949137.  
Last modified: 22 Nov 2009, 23:01:07 UTC

Looks like you're in a bunch of other projects too, which is perfectly fine, but it's all competition for resources. TDCF is one of those things that rises quickly, and corrects slowly as your system clears various hurdles.
ID: 949140 · Report as offensive
CodeRedDewd

Send message
Joined: 22 Nov 09
Posts: 5
Credit: 2,854
RAC: 0
United States
Message 949144 - Posted: 22 Nov 2009, 23:17:06 UTC - in response to Message 949140.  

I was only runny SETI and GPUGRID. SETI even stole the work from GPUGRID, so I was running 6 jobs of SETI. I Did the benchmark and restarted all the jobs. That was my Correction Factor.
ID: 949144 · Report as offensive
Profile 52 Aces
Avatar

Send message
Joined: 7 Jan 02
Posts: 497
Credit: 14,261,068
RAC: 67
United States
Message 949149 - Posted: 22 Nov 2009, 23:44:38 UTC - in response to Message 949144.  

Your TDCF should self correct. Check after a few hours.

Re-running benchmark can throw off the TDCF if CPU clocking was changed. But again, it should self correct in time. The nuisance short term impact is BOINC is probably throwing certain nearer deadline WU's into High Priority and is limiting the number of spare WU's in reserve.
ID: 949149 · Report as offensive
Richard Haselgrove Project Donor
Volunteer tester

Send message
Joined: 4 Jul 99
Posts: 14650
Credit: 200,643,578
RAC: 874
United Kingdom
Message 949150 - Posted: 22 Nov 2009, 23:46:22 UTC

Probably got a VLAR on his CUDA.
ID: 949150 · Report as offensive
Profile 52 Aces
Avatar

Send message
Joined: 7 Jan 02
Posts: 497
Credit: 14,261,068
RAC: 67
United States
Message 949152 - Posted: 22 Nov 2009, 23:53:43 UTC - in response to Message 949150.  

Probably got a VLAR on his CUDA.


I forgot about that case. Regardless, it looks like he just aborted all his tasks.
ID: 949152 · Report as offensive
CodeRedDewd

Send message
Joined: 22 Nov 09
Posts: 5
Credit: 2,854
RAC: 0
United States
Message 949178 - Posted: 23 Nov 2009, 3:48:42 UTC - in response to Message 949152.  

For some reason it kept crashing my computer, so I did abort the tasks... I'm still in the learning phases of this... I'm trying to find the tasks that run best on AMD on don't waste CPU time. It seems everyone has that question...
ID: 949178 · Report as offensive
CodeRedDewd

Send message
Joined: 22 Nov 09
Posts: 5
Credit: 2,854
RAC: 0
United States
Message 949180 - Posted: 23 Nov 2009, 3:53:54 UTC - in response to Message 949178.  

Claimed credit 174.626505732713
Granted credit 105.866692390517

It crunched on one core of my 9800GX2 for 3.17 hours and that's it? Is that because the correction factor hasn't corrected itself?

GPUGRID, I get about 300 per hour per GPU when I calculate with Run Time not CPU time.
ID: 949180 · Report as offensive
CodeRedDewd

Send message
Joined: 22 Nov 09
Posts: 5
Credit: 2,854
RAC: 0
United States
Message 949183 - Posted: 23 Nov 2009, 4:04:12 UTC - in response to Message 949180.  

After running more tasks it came down to 3.7583. I'm runing a huge task that originally said it would take 155 hours, and it is now at 23 with 16 left. I figure that's about 4 hours... What kind of credit shoud I expect? It seems like the factor would fix faster...
ID: 949183 · Report as offensive
1mp0£173
Volunteer tester

Send message
Joined: 3 Apr 99
Posts: 8423
Credit: 356,897
RAC: 0
United States
Message 949185 - Posted: 23 Nov 2009, 4:09:41 UTC - in response to Message 949137.  

It says that my Duration Correction Factor is 3.8286... What does that mean? From my understanding it means that my computer is much slower than what it predicts?

I have a Phenom 9950 @ 3.25Ghz 4gig ram and an XFX9800 GX2 overclocked as well. I thought my correction factor would be less than 1 for sure.

Don't overthink this.

The value is not exact. In fact, there are a lot of things that can throw it off.

BOINC tends to drive this number up alot faster than it falls, because a value that is too big doesn't hurt, while a value that is too small can make you miss deadlines.

Others have mentioned some more concrete reasons.
ID: 949185 · Report as offensive
1mp0£173
Volunteer tester

Send message
Joined: 3 Apr 99
Posts: 8423
Credit: 356,897
RAC: 0
United States
Message 949186 - Posted: 23 Nov 2009, 4:12:55 UTC - in response to Message 949183.  

After running more tasks it came down to 3.7583. I'm runing a huge task that originally said it would take 155 hours, and it is now at 23 with 16 left. I figure that's about 4 hours... What kind of credit shoud I expect? It seems like the factor would fix faster...

Let's say that BOINC predicted that a work unit would take four hours.

After 15 minutes the work unit finished because it was full of RFI.

If the DCF corrected very quickly, and you got a run of noisy work (which happens) then BOINC could decide that all four hour work units really took 15 minutes.

... and then, when we got a tape that wasn't noisy, it'd get a whole bunch of work units that it thought would take 15 minutes, but actually do take four hours.

That's bad.

To prevent that, DCF adjusts down very s-l-o-w-l-y.
ID: 949186 · Report as offensive
Luke
Volunteer developer
Avatar

Send message
Joined: 31 Dec 06
Posts: 2546
Credit: 817,560
RAC: 0
New Zealand
Message 949215 - Posted: 23 Nov 2009, 8:09:23 UTC

Your TDCF (Task Duration Correction Factor) is basically how fast/slow BOINC thinks it will take to complete a task.

A high TDCF (~1.50>) like yours indicates you are crunching those tasks 3.82 times quicker than BOINC thought you would, so as it lowers back to a normal level, BOINC will try and score you more tasks.

A low TDCF (<~0.80) indicates that you are crunching the tasks slower than BOINC thought you would, BOINC will compensate by requesting less work.

A TDCF of 1.02 to 0.98 shows BOINC is pretty spot on, in regards to how long it takes to complete said task.

- Luke.
- Luke.
ID: 949215 · Report as offensive
Profile Gundolf Jahn

Send message
Joined: 19 Sep 00
Posts: 3184
Credit: 446,358
RAC: 0
Germany
Message 949222 - Posted: 23 Nov 2009, 8:47:52 UTC - in response to Message 949180.  

Claimed credit 174.626505732713
Granted credit 105.866692390517

It crunched on one core of my 9800GX2 for 3.17 hours and that's it? Is that because the correction factor hasn't corrected itself?

GPUGRID, I get about 300 per hour per GPU when I calculate with Run Time not CPU time.

The DCF and the amount of credit granted are in no way correlated. DCF is solely used to calculate the time to completion.

Gruß,
Gundolf
Computer sind nicht alles im Leben. (Kleiner Scherz)

SETI@home classic workunits 3,758
SETI@home classic CPU time 66,520 hours
ID: 949222 · Report as offensive
FiveHamlet
Avatar

Send message
Joined: 5 Oct 99
Posts: 783
Credit: 32,638,578
RAC: 0
United Kingdom
Message 949229 - Posted: 23 Nov 2009, 10:54:08 UTC - in response to Message 949215.  

The DFC on my 2 main rigs is .3311 on the AMD and .2911 on the Intel i7.
Neither is in the range of 1.02 to .98.
I have had a problem on the AMD lately where Boinc decided that I needed 2000
plus tasks and in the download phase got screwed up and the S@H project dissappeared leaving the tasks.
I detached and then did a reatach and Boinc promply started downloading another 2000 plus tasks.
My question is if my DCF is so low why so many tasks.I do have Cuda but surely the DCF should account for that.

Dave
ID: 949229 · Report as offensive
Profile Gundolf Jahn

Send message
Joined: 19 Sep 00
Posts: 3184
Credit: 446,358
RAC: 0
Germany
Message 949239 - Posted: 23 Nov 2009, 13:58:13 UTC - in response to Message 949229.  

My question is if my DCF is so low why so many tasks.I do have Cuda but surely the DCF should account for that.

Your DCF is so low because you are running CUDA, which is faster than expected by BOINC (I think Luke got it the other way round).

If you think that BOINC downloads too many tasks, lower your cache settings.

Gruß,
Gundolf
Computer sind nicht alles im Leben. (Kleiner Scherz)

SETI@home classic workunits 3,758
SETI@home classic CPU time 66,520 hours
ID: 949239 · Report as offensive
Profile tullio
Volunteer tester

Send message
Joined: 9 Apr 04
Posts: 8797
Credit: 2,930,782
RAC: 1
Italy
Message 949242 - Posted: 23 Nov 2009, 14:12:24 UTC
Last modified: 23 Nov 2009, 14:27:21 UTC

I agree with Gundolf. My TDCF in AQUA is 35, in SETI is 0.157373. In my other BOINC projects is somewhat less than 1.
Tullio
ID: 949242 · Report as offensive
FiveHamlet
Avatar

Send message
Joined: 5 Oct 99
Posts: 783
Credit: 32,638,578
RAC: 0
United Kingdom
Message 949268 - Posted: 23 Nov 2009, 16:17:10 UTC

Thank's for the info both.

Dave
ID: 949268 · Report as offensive
Ingleside
Volunteer developer

Send message
Joined: 4 Feb 03
Posts: 1546
Credit: 15,832,022
RAC: 13
Norway
Message 949281 - Posted: 23 Nov 2009, 17:14:35 UTC - in response to Message 949215.  

Your TDCF (Task Duration Correction Factor) is basically how fast/slow BOINC thinks it will take to complete a task.

A high TDCF (~1.50>) like yours indicates you are crunching those tasks 3.82 times quicker than BOINC thought you would, so as it lowers back to a normal level, BOINC will try and score you more tasks.

A low TDCF (<~0.80) indicates that you are crunching the tasks slower than BOINC thought you would, BOINC will compensate by requesting less work.

A TDCF of 1.02 to 0.98 shows BOINC is pretty spot on, in regards to how long it takes to complete said task.

- Luke.

You've turned-it backwards. ;)

Basically, it is "Estimated cpu-times based on benchmarks and projects estimates FLOPS" * DCF = "actual cpu-time".

So, example if project-estimates is 10 hours, and DCF is 1.5, the computer has actually used 15 hours. Similarly, with the same 10-hour-estimate, a DCF of 0.1 means computer has used only 1 hour.

Preferably all DCF should be 0.9xxx, since this would indicate project got the estimates nearly correct, but they're a little on the high-side, so anyone just attaching won't get more work than can handle.

"I make so many mistakes. But then just think of all the mistakes I don't make, although I might."
ID: 949281 · Report as offensive

Message boards : Number crunching : Correction Factor of 3.8286???


 
©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.