Message boards :
Number crunching :
Credits : how is it calculated ???
Message board moderation
Author | Message |
---|---|
marsinph Send message Joined: 7 Apr 01 Posts: 172 Credit: 23,823,824 RAC: 0 |
On the same computer some WU runs faster but receive higher credits than other. Some runs slower and receive also higher credits ! Also for the same running time there are big difference. Of course for similar WU (same version) !!! I not compare WU running on GPU also not compare AP with other Who can explain ? |
rob smith Send message Joined: 7 Mar 03 Posts: 22199 Credit: 416,307,556 RAC: 380 |
So many folks ask that question...... Crudely, and this is the very simplest way of explaining it: The time taken by each of the two "valid" wingmen is normalised, and the lower time is used to calculate the credit to be awarded using some (very inaccurate) estimates of the "FLOPS" rate of each of the processors. (Both the "normalisation" and "FLOPS estimation" processes are prone to significant errors) Bob Smith Member of Seti PIPPS (Pluto is a Planet Protest Society) Somewhere in the (un)known Universe? |
rob smith Send message Joined: 7 Mar 03 Posts: 22199 Credit: 416,307,556 RAC: 380 |
AP are a totally different "breed" of task, but suffer the same issues...... Bob Smith Member of Seti PIPPS (Pluto is a Planet Protest Society) Somewhere in the (un)known Universe? |
marsinph Send message Joined: 7 Apr 01 Posts: 172 Credit: 23,823,824 RAC: 0 |
Tks Rob, but how is is calculated ? I understand about normalisation |
rob smith Send message Joined: 7 Mar 03 Posts: 22199 Credit: 416,307,556 RAC: 380 |
Once the data has been normalised it is translated from the time domain into FLOPS, then scaled to give the awarded credits. Both wingmen get the same amount of credit for a given task. The driver is the CPU time, with scaling to take into account the differences between processors. There are various bounds checks along the way to make sure that the end product is "within reason", and various so called "damping factors" are applied to try to reduce the effects of changes in processor and applications. Only the bounds checking works properly most of the time, the damping factors are not correct, and are gradually forcing the awarded credit towards zero (the timecale for which is years) Bob Smith Member of Seti PIPPS (Pluto is a Planet Protest Society) Somewhere in the (un)known Universe? |
Wiggo Send message Joined: 24 Jan 00 Posts: 34744 Credit: 261,360,520 RAC: 489 |
Oh just tell him the truth. It's just a random number generator that just gradually generates lower numbers. ;-) Cheers. |
juan BFP Send message Joined: 16 Mar 07 Posts: 9786 Credit: 572,710,851 RAC: 3,799 |
Oh just tell him the truth. That´s why we called Credit Screw in the real world. LOL |
rob smith Send message Joined: 7 Mar 03 Posts: 22199 Credit: 416,307,556 RAC: 380 |
A one word explanation of how it's calculated: Badly Bob Smith Member of Seti PIPPS (Pluto is a Planet Protest Society) Somewhere in the (un)known Universe? |
bluestar Send message Joined: 5 Sep 12 Posts: 7031 Credit: 2,084,789 RAC: 3 |
The more we are supposed to know, the better we also should be able to explain it. |
Keith Myers Send message Joined: 29 Apr 01 Posts: 13164 Credit: 1,160,866,277 RAC: 1,873 |
A one word explanation of how it's calculated: +1 Seti@Home classic workunits:20,676 CPU time:74,226 hours A proud member of the OFA (Old Farts Association) |
Richard Haselgrove Send message Joined: 4 Jul 99 Posts: 14650 Credit: 200,643,578 RAC: 874 |
The credit awarded is totally screwed up.True. But the credit calculations are made by, and the responsibility of, BOINC. SETI is one of the few projects which reveals the true problems with the raw BOINC credit system. You should direct your comments about resourcing to BOINC, not SETI. |
Richard Haselgrove Send message Joined: 4 Jul 99 Posts: 14650 Credit: 200,643,578 RAC: 874 |
3. Look on the bright side, you have advance knowledge in your other role :-))Indeed. I am still the only person to have commented on BOINC Server issue #2132 - and nobody has done any work on it since I captured those graphs in 2014. |
juan BFP Send message Joined: 16 Mar 07 Posts: 9786 Credit: 572,710,851 RAC: 3,799 |
I remember that and even participate on some of the tests and the conclusion was clear, Credit Screw is screwed. Maybe that explain why other projects like E@H not use Credit Screw. But as allways, the worst blind is the one who not want to see. |
arkayn Send message Joined: 14 May 99 Posts: 4438 Credit: 55,006,323 RAC: 0 |
We do have Richard on our side and long term posters such as he and I that are not giving up. But it is galling when Eric says that we need more computing power, and it is likely that if we sorted out this credit business we could get it!! Yep, crunch 8 minutes here and receive, maybe 100 points. Crunch another project 8 minutes and receive 1000, 10000, or 20000. Hmmmmmmmmmmmmm............ |
Grant (SSSF) Send message Joined: 19 Aug 99 Posts: 13736 Credit: 208,696,464 RAC: 304 |
Yep, crunch 8 minutes here and receive, maybe 100 points. Crunch another project 8 minutes and receive 1000, 10000, or 20000. Or 100,000. Grant Darwin NT |
juan BFP Send message Joined: 16 Mar 07 Posts: 9786 Credit: 572,710,851 RAC: 3,799 |
The worst side of credit screw is (ussing arkayn example) If you crunch for 8 min one type of WU you receive 100 credits (hypoteticaly) Next you crunch the same 8 min, on the same host a similar WU (with the same AR and type) and you could receive 50, 60, 80, or wath else the random credits generator gives. So it´s not just the low credit paid compared with other projects (with that i could easely live) , is the incapacity to paid the same amount of credit for the same crunching time on the same type of WU on the same host. Now if you compare the credit paid Arecibo vs blc vs AP then its a total mess. Just watch, almost everyone like to receive a lot of AP WU, since they paid a lot more credit per crunching time against the others, on the same host. As somebody allways say, credits mean nothing, you can´t exchange credits for nothing (not even the Seti Toaster) and science is done anyway (and that is true), but us homans like to compare our work done and the only way to measure that is with the credit production of the host. I belive very few knows where or how to discover the daily WU production (by WU type, AR, etc) on a determinate host//users, to make a real comparison. At least i don´t know an easy way. My 0.02Cents |
Grant (SSSF) Send message Joined: 19 Aug 99 Posts: 13736 Credit: 208,696,464 RAC: 304 |
So it´s not just the low credit paid compared with other projects (with that i could easely live) , is the incapacity to paid the same amount of credit for the same crunching time on the same type of WU on the same host. Yeah, I often raise my eyebrows when I notice a WU that took 5 minutes to crunch gets less Credit than one that took 7 min to crunch. But the worst are when you get 2 WUs with identical runtimes, and the Credit can vary by as much as 80 (when the larger amount awarded was only 150). Grant Darwin NT |
David@home Send message Joined: 16 Jan 03 Posts: 755 Credit: 5,040,916 RAC: 28 |
LOL please tell which projects give the best credit scores, I'll make one of them my backup project! |
betreger Send message Joined: 29 Jun 99 Posts: 11361 Credit: 29,581,041 RAC: 66 |
LOL please tell which projects give the best credit scores, I'll make one of them my backup project! Collatz probably pays the most but my problem with it is that It made me feel more stupid than normal after a couple of weeks. |
©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.