Message boards :
Number crunching :
Credit
Message board moderation
Author | Message |
---|---|
[B^S] madmac Send message Joined: 9 Feb 04 Posts: 1175 Credit: 4,754,897 RAC: 0 |
|
HAL9000 Send message Joined: 11 Sep 99 Posts: 6534 Credit: 196,805,888 RAC: 57 |
Shorties are less than average pay typically. SETI@home classic workunits: 93,865 CPU time: 863,447 hours Join the [url=http://tinyurl.com/8y46zvu]BP6/VP6 User Group[ |
Cliff Harding Send message Joined: 18 Aug 99 Posts: 1432 Credit: 110,967,840 RAC: 67 |
Has this changed just checked my account and for roughly 2 hours work getting only in 20's credit thought I got more for these shorties or am I wrong. Thanks in advance You think that's back, check this out. Granted that I went out with a -9 error, my wingman didn't. It still was validated on both. http://setiathome.berkeley.edu/workunit.php?wuid=767522396 |
-BeNt- Send message Joined: 17 Oct 99 Posts: 1234 Credit: 10,116,112 RAC: 0 |
Has this changed just checked my account and for roughly 2 hours work getting only in 20's credit thought I got more for these shorties or am I wrong. Thanks in advance I've been going through the credits granted per workunit I've done. If I divide up the CPU time by the credit (time/cr) most of my mb units have been between 50-58. Yours, on the fastest time there is 53.35 so that's normal credit amounts if you ask me. If anything the other guy got jipped with a 164.5 ratio! lol. Traveling through space at ~67,000mph! |
Bernie Vine Send message Joined: 26 May 99 Posts: 9954 Credit: 103,452,613 RAC: 328 |
|
Tim Send message Joined: 19 May 99 Posts: 211 Credit: 278,575,259 RAC: 0 |
I'd say that the CreditNew system is getting worse and and worse. We're getting less and less credit per hour crunching.[/quote] I agree. Even the best computers out there have difficulties to pass 100k. My rack also going down even i crunch 24/7 with 3 vgas. |
HAL9000 Send message Joined: 11 Sep 99 Posts: 6534 Credit: 196,805,888 RAC: 57 |
I can do better than that: Oh yeah, take this! :) http://setiathome.berkeley.edu/workunit.php?wuid=767612252 Actually I find it a bit odd. That I have two results on the same machine, with the same AR's, with similar run times(3,879.18s vs 3,875.84s) & with vastly different credit. Both were validated with the same wingman & app. http://setiathome.berkeley.edu/workunit.php?wuid=767804162 http://setiathome.berkeley.edu/workunit.php?wuid=767804283 That one defiantly makes me scratch my head there. SETI@home classic workunits: 93,865 CPU time: 863,447 hours Join the [url=http://tinyurl.com/8y46zvu]BP6/VP6 User Group[ |
rob smith Send message Joined: 7 Mar 03 Posts: 22237 Credit: 416,307,556 RAC: 380 |
Shorties will by the their very nature pay less per hour. Its quite simple really, there is a fixed overhead of a few seconds at the beginning and end of each WU. This overhead is pretty much constant for a particular cruncher, thus for a shorty it will be a bigger fraction of the total time than for a "longy". So if the overhead for a cruncher is 1 minute for each WU and a shorty takes 10 elapsed minutes and a longy takes 60 elapsed minutes you loose 6 minutes per hour doing shorties and only 1 minute per hour for longies. (The overhead is included in the apparent calculation time, and the times are to demonstrate the point, not to be representative of any given cruncher). So you will loose about ten percent credits hour for doing shorties over doing longies. (Note "elepased minutes" is the CPU time from starting to work on a WU to the time that WU is available to be reported and is the sum of non-productive overhead and productive calculation) Bob Smith Member of Seti PIPPS (Pluto is a Planet Protest Society) Somewhere in the (un)known Universe? |
Donald L. Johnson Send message Joined: 5 Aug 02 Posts: 8240 Credit: 14,654,533 RAC: 20 |
I can do better than that: And look at the stderr summaries. 767804283 returned many more signals than did 767804162, But 767804162 got waaaay more credit. Indeed, a puzzlement. Donald Infernal Optimist / Submariner, retired |
kittyman Send message Joined: 9 Jul 00 Posts: 51469 Credit: 1,018,363,574 RAC: 1,004 |
What can I tell ya? The ol' kittyman don't sweat the credit rates anymore. Credit new or credit old, whatever the system is that's in place, it treats the kitties the same as any other cruncher on the project. So, should the rate of credit granted on Seti be cut in half tomorrow, I DON'T CARE. Because I know that the cruncher ahead of me and the cruncher behind me are all in the same boat (along with that cute little ducky that likes to hop in for a ride now and then). Whatever my RAC on THIS PROJECT, it still gives me a valid comparison to anybody else doing Seti. And I don't care either, if I could triple my RAC by crunching XYZ@Home.....and I probably could. 'Cuz Seti be my binness. Meow! "Freedom is just Chaos, with better lighting." Alan Dean Foster |
Link Send message Joined: 18 Sep 03 Posts: 834 Credit: 1,807,369 RAC: 0 |
Shorties will by the their very nature pay less per hour. That still does not explain why I get for one shorty not much over 20Cr and for an other one with almost same crunching time and AR 65Cr. Here: AR=2.728446, 5,067.70 CPU seconds, 65.18 Credits AR=2.724110, 5,054.17 CPU seconds, 22.02 Credits |
Claggy Send message Joined: 5 Jul 99 Posts: 4654 Credit: 47,537,079 RAC: 4 |
Shorties will by the their very nature pay less per hour. The amount of Credit awarded with New Credit is based on Run Time and the Processing Rate of the app (as well as some other multipliers), If you're running Anonymous platform and your Wingman is running Stock, then it's always going to be your wingman that determines the Credit (unless he misses the deadline), Your wingman in this case has had a number of Download Error's, this might have bumped his Aveage Processing Rate up, because his GPU is supposible faster now, the Wu should finish quicker, since it doesn't, it just means in New Credits eye's it's a longer Wu, more runtime means more Credit, Claggy |
Speedy Send message Joined: 26 Jun 04 Posts: 1643 Credit: 12,921,799 RAC: 89 |
Workunit 768428323 gave 1.95 credit. I feel for the person that spent 0.89 of an hour on this task & received 1.95 credits for their work. My wu is the GPU unit. I'm using Multibeam x38g application & Boinc 6.10.58 Both unit are WU from tape 27fe11ag Wu true angle range is : 1.424575 & found the following Spike count: 9 Pulse count: 0 Triplet count: 3 Gaussian count: 0 |
Lint trap Send message Joined: 30 May 03 Posts: 871 Credit: 28,092,319 RAC: 0 |
Perhaps a graphical illustration... The middle two wu's are Fine, no problem with them, but notice the first and last wu. What's up with that??!! Wingmate for the 3rd wu used the cpu, the others were cuda of various vintages. Martin /edited/ |
rob smith Send message Joined: 7 Mar 03 Posts: 22237 Credit: 416,307,556 RAC: 380 |
It does look as if the query to calculate the credit isn't working correctly. From my understanding it should attempt to normalise the performance of the CPU/GPU. Then it calculates the credit to be granted. Just now it appears to be working with a strange mixture of normalised and un-normalised figures so sometimes you are "lucky" and get paired with a slower cruncher as lead, and get the more elevated score, and sometimes a fast cruncher as lead and get his reduced score. I don't think its a deliberate act, but a corrupt table - the one that is used to hold the data used to calculate the relative performance of the crunchers... (faster/slower here are compared to some mythical "standard" CPU/GPU as used for the normalisation process) ("lead" in this context is the "master" cruncher for the credit calculation, normally I think this is the first one on the list, but I don't think this is always the case) 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: 22237 Credit: 416,307,556 RAC: 380 |
I've just done a quick calculation on a few recent validated WU, using the declared formula: Credit = ([whetstone]+ [Dhrystone])/1000 * 100 / (2 * secs_per_day) * wu_cpu_time You get whetstones and dhrystones for a cpu by running the benchmarks, the rest is quite obvious as they come from the results for a particular cpu. Using this I calculated the results, and found that I was being "under paid" by between 10 and 30 credits for a lot of the recent shorty storm, but sometimes it was more or less correct... So, what's changed? The overall logic hasn't, so it can only be in two places, either the whetstones/dhrystones aren't being read correctly for all cpu, or the seconds per day is wrong. I think we can discount the second, as that would be uniform -all credits would be out by the same proportion; this leaves the cpu performance figures as being read incorrectly. We know from other threads that some tables have suffered "bit-rot" in the recent server woes, I wonder if this is another table that has been damaged in a minor (non-fatal) way, so needs to be re-created? So what is the effect of having a wrong whetstone or dhrystone value - a worked example (using rounded figures from one of my crunchers) First, using the 'stones in the S@H database cpu run time = 7500s Whetstones = 1100 Dhrystones = 4200 credit = (5300 / 1728000) * 7500 = 23 credits Now using the 'stones that that reported by a recent benchmark run: cpu run time = 7500s Whetstones = 2000 Dhrystones = 4200 credit = (6200 / 1728000) * 75000 = 27 credits. Most crunchers have small (+/- 10) changes between benchmark runs, but hundreds - that's something to keep our eyes on. And raises a question - how do these figures get fed back from the local cruncher to the server concerned, and how often? For an explanation of the credit calculation see: http://www.boinc-wiki.info/BOINC_FAQ:_Credit 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: 22237 Credit: 416,307,556 RAC: 380 |
In my last contribution I deliberately said "cpu", not "cpu/gpu". The reason being I can't find how to obtain the 'stones for an installed gpu. So, where can I see the 'stones for a gpu, and can I force their calculation? (And I do realise they will vary far more than those for a cpu given the dynamic nature of the clock speeds of many gpu). 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: 22237 Credit: 416,307,556 RAC: 380 |
Many thanks. Bob Smith Member of Seti PIPPS (Pluto is a Planet Protest Society) Somewhere in the (un)known Universe? |
Slavac Send message Joined: 27 Apr 11 Posts: 1932 Credit: 17,952,639 RAC: 0 |
Regardless of how much time I put into a task, I get an average of 25.00 credits for CUDA apps. :/ http://setiathome.berkeley.edu/results.php?userid=9475661&offset=0&show_names=0&state=3&appid= Executive Director GPU Users Group Inc. - brad@gpuug.org |
perryjay Send message Joined: 20 Aug 02 Posts: 3377 Credit: 20,676,751 RAC: 0 |
Welcome to the shorty storm Slavic. You do have a few normal angle range work units turned in lately that got ~80 to 120 credits so that's pretty much the way it's going. By the way, we can't see your link when you do it that way. It's easy enough to just click on your user name and go that way. PROUD MEMBER OF Team Starfire World BOINC |
©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.