Welcome to MilkyWay@home

Longer WU's for same Amount of Credit ???

Message boards : Number crunching : Longer WU's for same Amount of Credit ???
Message board moderation

To post messages, you must log in.

AuthorMessage
STE\/E

Send message
Joined: 29 Aug 07
Posts: 486
Credit: 576,516,217
RAC: 36,784
Message 37911 - Posted: 1 Apr 2010, 22:11:38 UTC

The Wu's have increased by about 1/3 but the Credit hasn't, right now Milkyway Credit is lower than Collatz or DNETC ... ???
STE\/E
ID: 37911 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile banditwolf
Avatar

Send message
Joined: 12 Nov 07
Posts: 2425
Credit: 524,164
RAC: 0
Message 37912 - Posted: 1 Apr 2010, 22:15:34 UTC - in response to Message 37911.  

The Wu's have increased by about 1/3 but the Credit hasn't, right now Milkyway Credit is lower than Collatz or DNETC ... ???

April Fools?? :(
Doesn't expecting the unexpected make the unexpected the expected?
If it makes sense, DON'T do it.
ID: 37912 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Donnie
Avatar

Send message
Joined: 19 Jul 08
Posts: 67
Credit: 272,086,462
RAC: 0
Message 37913 - Posted: 1 Apr 2010, 22:24:29 UTC - in response to Message 37911.  

The Wu's have increased by about 1/3 but the Credit hasn't, right now Milkyway Credit is lower than Collatz or DNETC ... ???


I'm seeing a 70% increase in time with no increase in credits. Guess it's time to go to Collatz. Cards run cooler on single precision anyay.
ID: 37913 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
JAMC

Send message
Joined: 9 Sep 08
Posts: 96
Credit: 336,443,946
RAC: 0
Message 37916 - Posted: 2 Apr 2010, 0:34:02 UTC - in response to Message 37913.  

The Wu's have increased by about 1/3 but the Credit hasn't, right now Milkyway Credit is lower than Collatz or DNETC ... ???


I'm seeing a 70% increase in time with no increase in credits. Guess it's time to go to Collatz. Cards run cooler on single precision anyay.


Yep, me too... Collatz bound with everything...
ID: 37916 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile David Glogau*
Avatar

Send message
Joined: 12 Aug 09
Posts: 172
Credit: 645,240,165
RAC: 0
Message 37924 - Posted: 2 Apr 2010, 1:14:00 UTC

Bye, have fun, see ya later.

NO wait, I am already have a presence there hahaha.

Don't worry team, I am here for the science, so I will stay!
Must admit, haven't seen the old nothing then 140% then count backwards routine for a few weeks now. ~2:37 on the 4 x aircooled 5970's and ~2:20 on the 2 x water cooled ones.
ID: 37924 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Donnie
Avatar

Send message
Joined: 19 Jul 08
Posts: 67
Credit: 272,086,462
RAC: 0
Message 37926 - Posted: 2 Apr 2010, 1:36:10 UTC - in response to Message 37924.  

Bye, have fun, see ya later.

NO wait, I am already have a presence there hahaha.

Don't worry team, I am here for the science, so I will stay!
Must admit, haven't seen the old nothing then 140% then count backwards routine for a few weeks now. ~2:37 on the 4 x aircooled 5970's and ~2:20 on the 2 x water cooled ones.


Lots of luck. Just checked your time and they've increased to 160 seconds from 108 seconds for the same credits. I believe that's a 48% reduction in effort. Don't take my word, just have a look.
ID: 37926 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Len LE/GE

Send message
Joined: 8 Feb 08
Posts: 261
Credit: 104,050,322
RAC: 0
Message 37928 - Posted: 2 Apr 2010, 1:53:34 UTC

Yep, seeing it here too.
Old WU times ~110s (3.28897e+013 floatingpoint ops on GPU, 2.47165e+008 on FPU)
New WU times ~161s (4.67516e+013 floatingpoint ops on GPU, 3.84156e+008 on FPU)
Same credit.
The new once should give ~305 instead of 213.
:(

First of April is over, so please correct this now.
ID: 37928 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile skildude
Avatar

Send message
Joined: 13 Jan 08
Posts: 52
Credit: 59,823,507
RAC: 0
Message 37931 - Posted: 2 Apr 2010, 2:41:51 UTC - in response to Message 37928.  
Last modified: 2 Apr 2010, 3:01:21 UTC

and we're getting the % going to 140%+ and then counting down. Looks like a weird bunch of WU's and now I cant send work back
perfect
It looks like it cleared
Blah blah blah you know the rest
ID: 37931 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
CTAPbIi

Send message
Joined: 4 Jan 10
Posts: 86
Credit: 51,753,924
RAC: 0
Message 37932 - Posted: 2 Apr 2010, 2:58:09 UTC
Last modified: 2 Apr 2010, 3:54:25 UTC

480sec against 327sec before, i.e 47% down

update: 4870 OC'd 850/900, crunching 2WUs simultaneously
ID: 37932 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile skildude
Avatar

Send message
Joined: 13 Jan 08
Posts: 52
Credit: 59,823,507
RAC: 0
Message 37933 - Posted: 2 Apr 2010, 3:01:43 UTC - in response to Message 37932.  

110 before now 155
Blah blah blah you know the rest
ID: 37933 · 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
Message 37935 - Posted: 2 Apr 2010, 4:21:15 UTC - in response to Message 37928.  

New WU times ~161s (4.67516e+013 floatingpoint ops on GPU, 3.84156e+008 on FPU)
Same credit.
The new once should give ~305 instead of 213.

To be correct, the current credit multiplier for double precision operations should be 5.44 credits per calculated Tflop (10^12 flop). That means the WU from your example is worth

46.7516 + 0.0004 TFlop (GPU + CPU)
=> 46.752 TFlop * 5.44 credits/DP-TFlop = 254.33 credits

The last round of optimizations in september last year reduced the amount of necessary operations a bit (hence the speedup around that time), but the project sticked to the old flops count and didn't adapted the new values.
ID: 37935 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
CTAPbIi

Send message
Joined: 4 Jan 10
Posts: 86
Credit: 51,753,924
RAC: 0
Message 37936 - Posted: 2 Apr 2010, 4:29:03 UTC

I restarted my rig, got new WUs, now it's OK. Probably, "bad" batch? :-)
ID: 37936 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
schizo1988

Send message
Joined: 26 Nov 08
Posts: 6
Credit: 44,903,350
RAC: 0
Message 37937 - Posted: 2 Apr 2010, 7:13:05 UTC


My jobs seem to have returned to their old run times or very close, so need to run away yet. It may be about earning credits but it should not be all about it, people seem to jump from project to project as soon as they get a few hours of lower production, usually it works itself out and if the discrepancy is big the project usually makes an adjustment before long
ID: 37937 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
STE\/E

Send message
Joined: 29 Aug 07
Posts: 486
Credit: 576,516,217
RAC: 36,784
Message 37940 - Posted: 2 Apr 2010, 8:20:44 UTC - in response to Message 37936.  

I restarted my rig, got new WUs, now it's OK. Probably, "bad" batch? :-)


Mine have returned to their normal Times too without rebooting but I think it may have had more to do with Travis messing with the Validator from what I read in another Thread, could be wrong on that though ...

STE\/E
ID: 37940 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
STE\/E

Send message
Joined: 29 Aug 07
Posts: 486
Credit: 576,516,217
RAC: 36,784
Message 37941 - Posted: 2 Apr 2010, 8:21:35 UTC - in response to Message 37935.  

New WU times ~161s (4.67516e+013 floatingpoint ops on GPU, 3.84156e+008 on FPU)
Same credit.
The new once should give ~305 instead of 213.

To be correct, the current credit multiplier for double precision operations should be 5.44 credits per calculated Tflop (10^12 flop). That means the WU from your example is worth

46.7516 + 0.0004 TFlop (GPU + CPU)
=> 46.752 TFlop * 5.44 credits/DP-TFlop = 254.33 credits

The last round of optimizations in september last year reduced the amount of necessary operations a bit (hence the speedup around that time), but the project sticked to the old flops count and didn't adapted the new values.


CP, I'm happy with the Credits I get get now but am I reading you right and your saying we should be actually getting more but the Project hasn't raised the Credits to the Level they should be ... ???


STE\/E
ID: 37941 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Len LE/GE

Send message
Joined: 8 Feb 08
Posts: 261
Credit: 104,050,322
RAC: 0
Message 37943 - Posted: 2 Apr 2010, 9:56:36 UTC

While I made a rough guess based on the credits for a 'normal' WU, he made a calculation based on the FLOP count of the example I gave.
His results:
The 'normal' WU gives a little bonus for the project hickups and the longer WU has nearly the same % as penalty.

ID: 37943 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote

Message boards : Number crunching : Longer WU's for same Amount of Credit ???

©2024 Astroinformatics Group