Message boards :
Number crunching :
Response to concerns regarding the new credit system.
Message board moderation
Previous · 1 · 2 · 3 · 4 · 5 · 6 · 7 . . . 18 · Next
Author | Message |
---|---|
Lord Tedric Send message Joined: 18 Jun 99 Posts: 204 Credit: 1,063,736 RAC: 0 |
valid point - taken onboard |
Steve @ SETI.USA Send message Joined: 5 Sep 04 Posts: 189 Credit: 1,016,797 RAC: 0 |
I rely don't get it. How can it be unfair if the credits are rewarded based on number of operations CPU has done? That must be ultimate in fairness if anything. Some of us (yes I used optimized clients and Trux (didn't know it was rely cheating :( ) will get a bit lower RAC but still the relative advantage of having more/faster computers will remain (smaller but on the other hand it will be harder to bridge). We are not complaining about the credits being based on the actual operations done. We are complaining that the granted credit is so much less for more work - or rather not commensurate with the additional time it takes to crunch the new enhanced WUs. Example: I could crunch a typical regular WU in about 25 minutes and claim 32 credits for it. An enhanced WU takes much longer, but only gets the same credit. Now, let me show you why those of us who care about stats are a wee bit upset with this... Our team, SETI.USA, is currently the #1 SETI team for credits/day. We are still a young team (14 mos old). We motivate ourselves to overtake the older established teams by forecasting the amount of time it will take us to surpass the other teams' total credit. SETI.Germany has the largest total credit of all SETI teams - more than double ours. They accumulated this credit under the prior system, which is now changing. Even though both our team and SETI.Germany will be affected by the lower credit granted per day under enhanced, it will now take us much longer to pass their total credit. Look here for the current projections - http://www.boincsynergy.com/stats/overtake.php?project=sah&team=115396 As you can see, we are currently projected to overtake SETI.Germany's total credit in approximately 781 days - a little over 2 years. With the credit changes being seen with enhanced, it will now take us probably 4-6 years to pass their credit - or longer. Now, I know that most of you don't care about our team goals. But, this same issue will affect everyone who cares about stats! And there are far more volunteers who are motivated by and care about their stats than many here will acknowledge. I am willing to see how this settles out over the next 2 weeks for now. But if there are no significant changes that will be made to the credit system to keep it inline with the old system, as Eric stated - and we actually see the impact on stats that some of us are expecting - maybe Berkeley should just consider resetting the stats like they did when we switched from classic to BOINC. You would still see your pre-enhanced stats along with post-enhanced, for instance. This is obviously not the best solution IMO, and that is why we would simply prefer that the credit adjustment multiplier be properly adjusted to keep average credit inline with what it has been in the past. Yes, RAC matters! Yes, stats matter! We get nothing else for our time and money, and stats are the main thing that keep a lot of people crunching long-term! http://www.setiusa.net |
Scott Brown Send message Joined: 5 Sep 00 Posts: 110 Credit: 59,739 RAC: 0 |
I think that this would likely have been the best solution. Keep in mind that optimization is not the only flaw that existed under the old system. The benchmarking was screwed up for different operating systems, too (especially Linux). Thus, the existing cobbelstones from SETI are largely meaningless when compared the new system. Enhanced is supposed to establish a truly "level palying field", but by incorporating the flawed inequailities of the previous credit system via inclusion of current scores, this will never be the case. |
Xaak Send message Joined: 22 May 99 Posts: 32 Credit: 22,636,357 RAC: 0 |
@Zaak Sorry, I misread your initial post. Slowest boxes: Pentium 630 running at 3.825ghz with an Asus P5ld2-VM board (Intel 945p chipset) 1 gb DDR2-667 (pc2-5300) Regular Seti wus average under 35 minutes 2 at a time Pentium 530 running at 3.3 ghz with a gigabyte 945p chipset motherboard 2 gb DDR2-533 (pc2-4300) @ DDR2-586 speed. Regular seti wus average around 40 minutes 2 at a time Both of those were previously running ATI chipset board with standard DDR (Asus P5-RD* series boards) and crunch times for both were over an hour/wu 2 at a time. Someone previously stated that DDR2 has much higher bandwidth, which is true, and seti appears to take full advantage of that. Plus, the 945/955 chipset also takes advantage of being able to run memory async very well too. XaaK |
Josef W. Segur Send message Joined: 30 Oct 99 Posts: 4504 Credit: 1,414,761 RAC: 0 |
That is not so. It was assumed that benchmarks would indicate how fast a machine could do science application work, so that a faster machine would get about the same credit for the same work using the benchmarks * time method. The _intent_ was equal credit for equal work, though obviously it didn't work as intended. Joe |
AxMi-24 Send message Joined: 12 Aug 99 Posts: 7 Credit: 12,060,218 RAC: 6 |
Steve Akers I get what you mean. We are competing (but on country basis with our neighbours) but still I think that just the knowledge that you are advancing faster than others is good enough stimulant. Of course I agree that reseting the scores would be probably best solution. That way we can have one classic, pre-enhanced and enhanced score. (not rely sure if this enhanced thing is only seti or BOINC). |
Xaak Send message Joined: 22 May 99 Posts: 32 Credit: 22,636,357 RAC: 0 |
Wow, I never realized the entire seti project exists just to allow Seti.usa to pass other teams faster. And here I thought I was crunching for a science project ;-). XaaK |
Xaak Send message Joined: 22 May 99 Posts: 32 Credit: 22,636,357 RAC: 0 |
There's no need to factor in benchmarks in unless the speed of a machine is a factor in how many credits per wu are granted. The Enhanced system grants fixed credit per wu, theoretically at least, and benchmarks play no part in that calculation. XaaK |
Steve @ SETI.USA Send message Joined: 5 Sep 04 Posts: 189 Credit: 1,016,797 RAC: 0 |
Wow, I never realized the entire seti project exists just to allow Seti.usa to pass other teams faster. And here I thought I was crunching for a science project ;-). Xaak - I guess you missed the point. I was using our team merely as an example. This issue affects all teams and all volunteers, in case you missed that. And, it's not about moving faster - just moving at the same speed as before, for EVERYONE! http://www.setiusa.net |
Kango Send message Joined: 29 May 99 Posts: 6 Credit: 3,585,750 RAC: 0 |
FYI, Xaak. It's called an example. |
Mr.Pernod Send message Joined: 8 Feb 04 Posts: 350 Credit: 1,015,988 RAC: 0 |
funny how this "I have some spare cpu-cycles, let me help science" turned into "gimme my credit or else...." |
Andy Lee Robinson Send message Joined: 8 Dec 05 Posts: 630 Credit: 59,973,836 RAC: 0 |
It is all relative, everyone is now on a more level playing field, so we can all compete on machine merit, not on programming tweaks and hacks. Apps should be optimised for their respective platforms, and a fair credit allocated for the work, but conning the client that a 4 gflop machine is 22 gflops is in my opinion not fair if the same rationale is not applied to everyone. We can all make justifications. Yes, SSE3 is great, and optimisations should be used if possible, but I think it was a flaw in the original 4.07/4.18 planning of credit allocation not to be able to reconcile optimised apps vs benchmark and resulting in apple/pear comparisons. Perhaps the solution is intractable, but surely the best way after all is to allocate credit on fixed quota per WU. Those with optimised apps can do more, those with faster cpus and infrastructure can do more and gain more credit. If all apps are supplied optimised for each platform, then the playing field is level and incorruptible. The whole credit system is ruined beyond repair anyway because of the asymmetry introduced by 4.07/4.18 optimisations, and the gains achieved by those corrupting the system will unlikely to be reversed, unless there could be some way of reappraising the entire system and reallocating all credits in one sweep, but I suspect all the historical results have been deleted now. This seems to amount to petty willy waving by anoraks getting too involved with getting to be the 'best' in something that they can actually make an impact with, at the expense of having a real life by taking the credit system far too seriously and losing perspective on the real objectives - to do science. Yes, I'm guilty too! It is addictive, and I have spent far too much time on this at the expense of doing real work... but then I'm a sad anorak too :) Why not just scrap the entire credit system, and publish more REAL results, like what was found, statistics on best candidate signals and theories etc. There is still a huge amount of info in these signals that has no intelligent origin but still containing valuable data on interesting phenomena out there. I would like much more feedback on the results of analysis instead of petty squabbling about credits. This would be far more worthy of spending our valuable time on, the science and exploration. Just some thoughts, Andy. eg. Owner guess... Created 4 Jul 2004 14:14:49 UTC Total Credit 250,053.19 Recent average credit 2,881.47 CPU type GenuineIntel Intel(R) Pentium(R) D CPU 3.20GHz Number of CPUs 2 Operating System Microsoft Windows 2003 Professional Edition, Service Pack 1, (05.02.3790.00) Memory 1022.07 MB Cache 976.56 KB Measured floating point speed 21932.13 million ops/sec Measured integer speed 6364.57 million ops/sec Average upload rate 0.68 KB/sec Average download rate 162.69 KB/sec Average turnaround time 2.41 days Maximum daily WU quota per CPU 100/day (is this actually ever honoured?) Results 756
|
Jord Send message Joined: 9 Jun 99 Posts: 15184 Credit: 4,362,181 RAC: 3 |
maybe Berkeley should just consider resetting the stats like they did when we switched from classic to BOINC. Oh yes, I look forward to seeing everyone come post at the Q&P forums for anytime between 3 times the 4 and 55 days (deadlines on results times the quorum needed, thus the "about time" for credit to be granted) asking "why can't I post on the main forums?" ... well, you need credit to be able to post here. Without it, no posting in NC, the Cafe or Science. So you want to cripple everyone's ability to post here just to reset the chances of overtaking team A or team B? |
Scott Brown Send message Joined: 5 Sep 00 Posts: 110 Credit: 59,739 RAC: 0 |
maybe Berkeley should just consider resetting the stats like they did when we switched from classic to BOINC. What??? Are you kidding? The forum posting requirements are largely irrelevant to this discussion (and could easily be fixed by using pre-enhanced or enhanced credit for posting eligibility after a reset). The point is that, EVEN IGNORING THE OPTIMIZED CLIENT ISSUES, the credit system was so FUBAR from the beginning that a complete reset is necessary to get anything remotely close to a "level playing field". |
1mp0£173 Send message Joined: 3 Apr 99 Posts: 8423 Credit: 356,897 RAC: 0 |
This implies that the project could have blocked modified clients somehow. They could have, by simply not releasing the BOINC source, but means we'd never see BOINC for anything but a "mainstream" OS, or something that is 100% binary compatible. ... and if they did start trying to block "improved" BOINC clients, those clients could just report standard versions -- a "stealth" enhancement. "Acceptance" means a whole lot less than one might think -- blocking enhanced BOINC clients may not be impossible, but it is impractical. |
HFB1217 Send message Joined: 25 Dec 05 Posts: 102 Credit: 9,424,572 RAC: 0 |
Regardless of the date posted here I have been a member of Seti for almost 7 years. A four or Five hour E-WU gives me a max of 50 credits per CPU. This is an actual return. In the time of 5 hours with a regular seti WU I could do 10 WU per CPU for an average of 20 points ea WU for a total of 200 credits. The awards were not the highest nor the lowest. The median credit would have been approximately the same if I ran optimized or not. Because if I had not optimized it would have been the lowest and thrown out and if it was the highest it would like wise be discarded. I was very rarely if ever given the points requested by my computers. So I don't see where your statement of a 10% difference holds any water or is factual. Besides that I have over 30 hrs of work that was completed and the return was 0 because of validation error. They were completed and rejected through no fault of ours. You admited that one of the spliters was issuing defective work units. They are giving EH@home credits for the same problem as Seti@home had. I am some what annoyed as to how you are conducting this circus act. I will continue to process WU regardless and in spite of the way Berkeley has conducted itself. But do so rather unhappily. -- ****aka The WIZARD **** A Founding member Seti BBR Team Starfire**** Come and Visit Us at BBR TeamStarFire ****My 9th year of Seti****A Founding Member of the Original Seti Team Starfire at Broadband Reports.com **** |
SargeD@SETI.USA Send message Joined: 24 Nov 02 Posts: 957 Credit: 3,848,754 RAC: 0 |
Not true. When the optimized client first came out, the majority of people I know took a wait and see attitude. Since the "project" did not voice displeasure about the client after a time many people started switching to it. If they had even hinted that they did not approve of it most of us now using it would have never started. It was a matter of discussion on many teams message boards that we wanted to see if Berkeley approved or disapproved. |
1mp0£173 Send message Joined: 3 Apr 99 Posts: 8423 Credit: 356,897 RAC: 0 |
If you run BOINC 5.4.9 and SETI Enhanced, your claimed credit and granted credit will match. The only exception is when the majority of a quorum are running older versions of BOINC. |
HFB1217 Send message Joined: 25 Dec 05 Posts: 102 Credit: 9,424,572 RAC: 0 |
If you run BOINC 5.4.9 and SETI Enhanced, your claimed credit and granted credit will match. Please read more carefully the statements about credit awards were in reference to Optimized Seti from Trux and or Cruncher3. It has nothing about Enhanced points except to express a contradiction to what Berkeley stated about the difference in point awards was only 10%. Come and Visit Us at BBR TeamStarFire ****My 9th year of Seti****A Founding Member of the Original Seti Team Starfire at Broadband Reports.com **** |
1mp0£173 Send message Joined: 3 Apr 99 Posts: 8423 Credit: 356,897 RAC: 0 |
Please read more carefully the statements about credit awards were in reference to Optimized Seti from Trux and or Cruncher3. You may want to re-read the original statement from Eric Korpela as well. He stated that Enhanced credits are within 10% of standard 4.x clients, not within 10% of optimized. |
©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.