PDA

View Full Version : calculating credit



Clank [MM]
10-02-13, 01:26 PM
I know some have seen this, worth posting. So does this mean everytime a new baseline machine is used credit earned goes down?

http://boinc.berkeley.edu/wiki/Computation_credit

Also, sorry, did this from my phone and did not notice I posted in games.

DrPop
10-02-13, 01:30 PM
Well . . . I think that depends on if the project uses the "Credit New" crap or not to generate a "baseline" of credit. Some do and THANKFULLY some don't and just have a fixed credit per WU returned - which is how it should be, regardless of how fast you crunched that WU. So, newer, faster hardware gets more credit per day. Duh - but leave it to academics to screw that one up! :P

Slicker
10-02-13, 02:53 PM
The issue is that the original baseline computer listed which got 1000 Megaflops per day was a PIII 800 Mhz machine. The problem is that as newer CPUs came out with enhanced instruction sets and the apps were modified to make use of those new instructions, they actually got more than 200 credits per gigaflop per day. So, DA and his team decided that getting more credit for doing more work was unfair. He then came up with his "CreditNew" algorithm which attempts to give the exact same credit for a given CPU and time period on all projects. It would work if we all ran SETI and only SETI. But, since no other project uses the exact same percent of floating point calculations as SETI, it is still apples and oranges. If you don't believe me, downclock the memory on your GPU by 50% on a project using CreditNew where the memory has an impact and see what happens. You credit will go down a little, but it will also drag down the average for all others using that GPU which means they get less credit as well because they are all supposed to be the same. Don't even get me started on projects that use 50% GPU load and get the same credit awarded by CreditNew as projects where the GPU is at 99% load.

Bryan
10-02-13, 03:04 PM
But Slicker you are SO wrong. Remember that GPUs don't crunch on their own and must use CPUs therefore they aren't "efficient" and shouldn't be awarded big credits!

c303a
10-02-13, 03:06 PM
Don't even get me started on projects that use 50% GPU load and get the same credit awarded by CreditNew as projects where the GPU is at 99% load.
Come on Slicker. We're all friends here. Tell us what you really think. I'll bet it is the same as just about everyone else here.:mad::mad::mad::rolleyes::rolleyes::rolleyes: :D:D:D

DrPop
10-02-13, 04:34 PM
... So, DA and his team decided that getting more credit for doing more work was unfair. . .

And thus . . . the Kingdom of BOINC Land slipped into its Marxist underworld . . . where one man's P4 is equal to another's i7. "After all," mused the Rulers, "they are both in the same size case, they must do the same work!"

trigggl
10-02-13, 05:03 PM
I had a P3 that ran at 1GHz with Rambus memory. I wonder what that would do to credit new. I think the Rambus was at 800MHz. So what if it's serial?

Slicker
10-03-13, 09:44 AM
I had a P3 that ran at 1GHz with Rambus memory. I wonder what that would do to credit new. I think the Rambus was at 800MHz. So what if it's serial?
Since no one is running those old P3's any more, no one really knows whether the credits are screwed up now or not. But, given that overclocking is penalized and underclocking is rewarded the masses will revolt at some point -- unless they are the same masses that keep voting for the same politicians who make the same false promises every couple years.

Robert7
10-03-13, 11:27 AM
well said

pinhodecarlos
10-03-13, 11:40 AM
Do you guys want to run something useful instead of running for the credits? I could need some help on my math projects...lol

Mumps
10-03-13, 11:53 AM
Sorry Slicker, but I am still running a PIII system. Albeit it a Dual 1.3Ghz one, but PIII all the same. :)

Starting to have problems finding projects to attach it to though. Most of the projects I want to run are using capabilities that don't exist in a PIII in their default App. Like SSE, etc..

DrPop
10-03-13, 12:13 PM
. . .the masses will revolt at some point -- unless they are the same masses that keep voting for the same politicians who make the same false promises every couple years.

VERY good point. But seeing how I lack even a vote in the process of how BOINC awards credits, how do you suggest we revolt - without completely leaving the hobby? This is not a rhetorical question. Perhaps one answer is: to crunch only those projects that refuse to use Credit New or other similar Shenanigans? Any other good ideas for us to try?

Clank [MM]
10-03-13, 01:22 PM
I see issues where new crunchers get discouraged when seeing the numbers that old school crunchers have. Ya you can throw money at video cards but someone new will never be able to post stats at our level. It could hurt the over all growth in the end.

Clank [MM]
10-03-13, 01:55 PM
Sorry Slicker, but I am still running a PIII system. Albeit it a Dual 1.3Ghz one, but PIII all the same. :)

Starting to have problems finding projects to attach it to though. Most of the projects I want to run are using capabilities that don't exist in a PIII in their default App. Like SSE, etc..

Now I understand how you were cranking out the credits at Pirates........

Bryan
10-03-13, 03:56 PM
VERY good point. But seeing how I lack even a vote in the process of how BOINC awards credits, how do you suggest we revolt - without completely leaving the hobby? This is not a rhetorical question. Perhaps one answer is: to crunch only those projects that refuse to use Credit New or other similar Shenanigans? Any other good ideas for us to try?

No, there are always enough idiots willing to crunch whatever is out there. I've seen multiple projects where the crunchers complain about low credits. Then there is the group that comes along saying it is okay and they are in it for the science not the credits (I guess that makes them superior somehow). So in the end, projects like Constellation and CAS continue to pay crap with respect to other projects.