New Credit Adjustment?

Message boards : Number crunching : New Credit Adjustment?
Message board moderation

To post messages, you must log in.

1 · 2 · 3 · 4 . . . 17 · Next

AuthorMessage
Andre Howard
Volunteer tester
Avatar

Send message
Joined: 16 May 99
Posts: 124
Credit: 217,463,217
RAC: 0
United States
Message 787179 - Posted: 26 Jul 2008, 2:36:43 UTC

Is anyone else experiencing a reduction in credit within the last couple of hours? All of a sudden my 3 machines are all underclaiming against my wingmen by an average of .50 credits. All machines are running v8.

ID: 787179 · Report as offensive
Profile jason_gee
Volunteer developer
Volunteer tester
Avatar

Send message
Joined: 24 Nov 06
Posts: 7489
Credit: 91,093,184
RAC: 0
Australia
Message 787187 - Posted: 26 Jul 2008, 2:48:50 UTC
Last modified: 26 Jul 2008, 2:49:30 UTC

It gets Freakier :O

here is one of yours, where both yourself and wingman are running ak_v8 SSSE3x Win32 Build 41:

http://setiathome.berkeley.edu/workunit.php?wuid=303248166

Both report the same flops, and multiplier ...yet the claim shows different... something screwy or purposely changed at the server end I'd suggest.

Jason
"Living by the wisdom of computer science doesn't sound so bad after all. And unlike most advice, it's backed up by proofs." -- Algorithms to live by: The computer science of human decisions.
ID: 787187 · Report as offensive
Dr Who Fan
Volunteer tester
Avatar

Send message
Joined: 8 Jan 01
Posts: 3346
Credit: 715,342
RAC: 4
United States
Message 787194 - Posted: 26 Jul 2008, 3:00:28 UTC

I also have one that fits in the same category:
http://setiathome.berkeley.edu/workunit.php?wuid=304737461
I used AK V8 and claimed 19.11 - my "Anonymous" wingman used stock app and claimed 18.59, difference of .52 credits. Both machines returned 4 spikes.

ID: 787194 · Report as offensive
Profile JDWhale
Volunteer tester
Avatar

Send message
Joined: 6 Apr 99
Posts: 921
Credit: 21,935,817
RAC: 3
United States
Message 787208 - Posted: 26 Jul 2008, 3:33:02 UTC

Just checked... discovered same behavior here...

wuid=303990264

Maybe another "side effect" of the AstroPulse enabling changes?
ID: 787208 · Report as offensive
Andre Howard
Volunteer tester
Avatar

Send message
Joined: 16 May 99
Posts: 124
Credit: 217,463,217
RAC: 0
United States
Message 787211 - Posted: 26 Jul 2008, 3:41:15 UTC

Looks like Jason got it right, the server page is showing some problems. Think I'll just wait till its fixed before I send any more workunits in so I don't cost people credit.

ID: 787211 · Report as offensive
Terror Australis
Volunteer tester

Send message
Joined: 14 Feb 04
Posts: 1817
Credit: 262,693,308
RAC: 44
Australia
Message 787225 - Posted: 26 Jul 2008, 4:07:02 UTC

Going back through my latest results I've found up to 10 credits difference in the claim between myself and my wingman for some units. Sometimes I'm the high one sometimes I'm the low one (about 50/50).

There appears to be no consistency between CPU type, client version or operating system as to which is higher or lower. One thing I did notice was significant differences in the Flop counter after the first four places.

Maybe "line noise" or similar is randomly affecting some uploads from the server so that the unit requires a little bit of extra crunching for the client to make sense of it ?

Brodo
ID: 787225 · Report as offensive
Eric Korpela Project Donor
Volunteer moderator
Project administrator
Project developer
Project scientist
Avatar

Send message
Joined: 3 Apr 99
Posts: 1382
Credit: 54,506,847
RAC: 60
United States
Message 787231 - Posted: 26 Jul 2008, 4:34:03 UTC

There is a modification to the scheduler that just went into effect this afternoon to try to adjust credits better to match benchmark scores. The differences here are because one result was returned with the old scheduler and one with the new one. The differences should be less noticeable after this point.

Eric
@SETIEric@qoto.org (Mastodon)

ID: 787231 · Report as offensive
Alinator
Volunteer tester

Send message
Joined: 19 Apr 05
Posts: 4178
Credit: 4,647,982
RAC: 0
United States
Message 787288 - Posted: 26 Jul 2008, 5:08:12 UTC
Last modified: 26 Jul 2008, 5:09:17 UTC

Hmmm...

I assume you're talking about the benchmark scores for the theoretical reference machine?

Alinator
ID: 787288 · Report as offensive
W-K 666 Project Donor
Volunteer tester

Send message
Joined: 18 May 99
Posts: 19402
Credit: 40,757,560
RAC: 67
United Kingdom
Message 787315 - Posted: 26 Jul 2008, 5:22:27 UTC

Why, I have to ask, is why anyone would want to compare credits, or anything else, to benchmark scores.

As I pointed out on Beta, two computers, a Q6600 and a Athlon X2 6000+ on same OS, have similar benchmark scores BUT the Q6600 completes an AP unit in under half the time of the AMD. Similar performance is also observed doing MB units.
ID: 787315 · Report as offensive
Profile bounty.hunter
Volunteer tester
Avatar

Send message
Joined: 22 Mar 04
Posts: 442
Credit: 459,063
RAC: 0
India
Message 787382 - Posted: 26 Jul 2008, 6:02:25 UTC

Isn't Astropulse based on benchmarks and not credit multipliers ?

It's been some time since I've been on Beta so correct me if I'm wrong...


I assume the new scheduler should be set to only Astropulse and not MB...

That might be causing the problem....
ID: 787382 · Report as offensive
W-K 666 Project Donor
Volunteer tester

Send message
Joined: 18 May 99
Posts: 19402
Credit: 40,757,560
RAC: 67
United Kingdom
Message 787388 - Posted: 26 Jul 2008, 6:16:20 UTC - in response to Message 787382.  

Isn't Astropulse based on benchmarks and not credit multipliers ?

It's been some time since I've been on Beta so correct me if I'm wrong...


I assume the new scheduler should be set to only Astropulse and not MB...

That might be causing the problem....

The latest AP app and the previous one have credits calculated on FLops. But due to the way AP works, i.e. total FLops is known, this basically means the same fixed credits for all AP units.
ID: 787388 · Report as offensive
Profile littlegreenmanfrommars
Volunteer tester
Avatar

Send message
Joined: 28 Jan 06
Posts: 1410
Credit: 934,158
RAC: 0
Australia
Message 787478 - Posted: 26 Jul 2008, 13:18:38 UTC
Last modified: 26 Jul 2008, 13:18:53 UTC

Some of the differences mentioned here are piffling compared to the over 1,400 my machine claimed, when awarded only 700-odd. Now THAT'S a big diff!
(Of course, we're all in it for the science, so I'm at least prepared to "sacrifice" a few credits). after all, it's not like they actually have a tradeable value, is it?
ID: 787478 · Report as offensive
UncleVom

Send message
Joined: 25 Dec 99
Posts: 123
Credit: 5,734,294
RAC: 0
Canada
Message 787588 - Posted: 26 Jul 2008, 16:45:17 UTC - in response to Message 787478.  

Some of the differences mentioned here are piffling compared to the over 1,400 my machine claimed, when awarded only 700-odd. Now THAT'S a big diff!
(Of course, we're all in it for the science, so I'm at least prepared to "sacrifice" a few credits). after all, it's not like they actually have a tradeable value, is it?


The existing BOINC benchmark in the view of many here is completely bogus when comparing actual work completed over various processors and operating systems, which is why I believe the project went to counting flops.

An adjustment to better even out the MB credits over differing work unit angles would be good and fair.

The cynical side of me thinks it is more likely an attempt to bring MB credit to the level of AP, on which I have previously stated my views in the now sticky Astropulse Rumor thread.


UncleVom


ID: 787588 · Report as offensive
Profile Dirk Sadowski
Volunteer tester

Send message
Joined: 6 Apr 07
Posts: 7105
Credit: 147,663,825
RAC: 5
Germany
Message 787604 - Posted: 26 Jul 2008, 17:38:23 UTC


I understood it well enough..?
Now the benchmarks are back again in the calculation of the Credits?
So it's no longer recommened to use Crunch3r's BOINC V6.1.0 client with little bit higher benchmark-results?

ID: 787604 · Report as offensive
Profile jason_gee
Volunteer developer
Volunteer tester
Avatar

Send message
Joined: 24 Nov 06
Posts: 7489
Credit: 91,093,184
RAC: 0
Australia
Message 787612 - Posted: 26 Jul 2008, 17:52:52 UTC - in response to Message 787604.  
Last modified: 26 Jul 2008, 17:58:14 UTC


I understood it well enough..?
Now the benchmarks are back again in the calculation of the Credits?
So it's no longer recommened to use Crunch3r's BOINC V6.1.0 client with little bit higher benchmark-results?


I suspect it won't make a bit of difference what the benchmark says [at your end], as the server still uses flopcount for the calculation, but was probably tweaked slightly to bring (server end) calculation to closer reflect the 'benchmark system', meaning a 'reference PC benchmark' rather than redefining the credit calculation process. [That's how I interpret Eric's post anyway...]

Jason
"Living by the wisdom of computer science doesn't sound so bad after all. And unlike most advice, it's backed up by proofs." -- Algorithms to live by: The computer science of human decisions.
ID: 787612 · Report as offensive
Profile Dirk Sadowski
Volunteer tester

Send message
Joined: 6 Apr 07
Posts: 7105
Credit: 147,663,825
RAC: 5
Germany
Message 787673 - Posted: 26 Jul 2008, 21:27:52 UTC
Last modified: 26 Jul 2008, 21:28:39 UTC

My AK v8.0 result was compared with stock app and I got 2.2 less credits.. :-(
wuid=300706526
Uhh.. what's going on?
ID: 787673 · Report as offensive
Profile jason_gee
Volunteer developer
Volunteer tester
Avatar

Send message
Joined: 24 Nov 06
Posts: 7489
Credit: 91,093,184
RAC: 0
Australia
Message 787678 - Posted: 26 Jul 2008, 21:31:43 UTC - in response to Message 787673.  
Last modified: 26 Jul 2008, 21:32:13 UTC

My AK v8.0 result was compared with stock app and I got 2.2 less credits.. :-(
wuid=300706526
Uhh.. what's going on?

Yep, looks like one reported before the change, and one after again, as described in Eric's post. This time your claim is higher ... go figure
"Living by the wisdom of computer science doesn't sound so bad after all. And unlike most advice, it's backed up by proofs." -- Algorithms to live by: The computer science of human decisions.
ID: 787678 · Report as offensive
1mp0£173
Volunteer tester

Send message
Joined: 3 Apr 99
Posts: 8423
Credit: 356,897
RAC: 0
United States
Message 787694 - Posted: 26 Jul 2008, 21:51:02 UTC - in response to Message 787604.  


I understood it well enough..?
Now the benchmarks are back again in the calculation of the Credits?
So it's no longer recommened to use Crunch3r's BOINC V6.1.0 client with little bit higher benchmark-results?

Benchmark could mean "the benchmark routines in BOINC" but that's unlikely because we all know how well that worked. Glad it's gone.

"Benchmark" can also mean "the standard we measure against" -- the hypothetical 100-cobblestone machine is a "benchmark."

So, they've tweaked the multipliers a tiny bit to keep everything comparable.

If they'd changed back to Benchmark * Time, the credits would be wildly different across different CPUs, not just a few percent.
ID: 787694 · Report as offensive
Profile DaBrat and DaBear

Send message
Joined: 13 Dec 00
Posts: 69
Credit: 191,564
RAC: 0
United States
Message 787759 - Posted: 27 Jul 2008, 0:49:19 UTC
Last modified: 27 Jul 2008, 1:21:11 UTC

I'm not sure if this is relevant but I can tell you that even my smal units. 16+ are returning a,mot half a point less than those they are being compared with and this seems to be consistant.. What I return as 16.21 is being reported by the other person as 16.67.... hope I'm not killing someone else's credit. My 18.77 is returned as 19.30 by my wingman

Larger WUs have larger discrepencies...see below
928970335 4454214 24 Jul 2008 17:57:34 UTC 26 Jul 2008 8:38:01 UTC Over Success Done 18,076.27 49.30 49.30
928970336 3875073 24 Jul 2008 17:57:25 UTC 25 Jul 2008 8:04:21 UTC Over Success Done 20,799.38 50.69 49.30

Could it be the task duration correction factor is incorrect due to the last CPU benchmarks run? Each of my compos were showing at least 50million ops per second slower. When I reran benchmarks the task duration correction factor corrected itself.... We will see.
The above is a full 1.39 lower than my wingman's. Seems the larger the WU the larger the discrepancy.
ID: 787759 · Report as offensive
Profile DaBrat and DaBear

Send message
Joined: 13 Dec 00
Posts: 69
Credit: 191,564
RAC: 0
United States
Message 787771 - Posted: 27 Jul 2008, 1:50:03 UTC
Last modified: 27 Jul 2008, 1:57:56 UTC

Only ran one WU on the comp that was consistantly lower. This time my comp had the higher return as compared to my wingman. Think those CPU benchmarks have something to do with it. If you're like me and never shut down, or load BOINC at as soon as you boot while other system processes are loading, it can throw off the speed SETI sees your comp crunching at. I guess it can fluctuate.
928436480 3259798 24 Jul 2008 7:56:15 UTC 26 Jul 2008 9:04:16 UTC Over Success Done 10,639.28 71.84 71.84
928436479 4449925 24 Jul 2008 7:56:08 UTC 27 Jul 2008 1:46:49 UTC Over Success Done 6,073.87 73.41 71.84

UPDATE: Uploaded one WU from each comp and it seems the results are consistant. Even though one unit had a narrower margin it came back even with the wingman this time compared to the past day's results.
ID: 787771 · Report as offensive
1 · 2 · 3 · 4 . . . 17 · Next

Message boards : Number crunching : New Credit Adjustment?


 
©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.