Credit lowering
Message boards :
Number crunching :
Credit lowering
Message board moderation
Previous · 1 . . . 5 · 6 · 7 · 8 · 9 · Next
| Author | Message |
|---|---|
 banditwolf Send message Joined: 12 Nov 07 Posts: 2425 Credit: 524,164 RAC: 0  
|
By utilising CPs latest ap, you are getting MORE credit than you were before. Not with the sse2 app, only ~8-9% faster. Good but doesn't compensate. Doesn't expecting the unexpected make the unexpected the expected? If it makes sense, DON'T do it. |
Bruce Send message Joined: 28 Apr 08 Posts: 1415 Credit: 2,716,428 RAC: 0  
|
By utilising CPs latest ap, you are getting MORE credit than you were before. I agree, I installed opp app on both my machines and watched my rac drop like a rock on these machines. One went from aprox 1200 to 950, the other went from 2800 to 2200, not exactly a plus??These are CPU machines. My rac has gone up but thats because I bought an i7 also CPU on MW. MY GPU's are at collatz where they get a LOT more than they would at MW. Cuda nVidia cards. |
The Gas Giant Send message Joined: 24 Dec 07 Posts: 1947 Credit: 240,884,648 RAC: 0 
|
Fair enough. I wasn't thinking about cpu's. My bad. But it is still more than what you'd get on SETI with a lunatics opt ap..... |
|
Send message Joined: 21 Aug 08 Posts: 625 Credit: 558,425 RAC: 0
|
By utilising CPs latest ap, you are getting MORE credit than you were before. Agreed... my Pentium 4 Before the reduction the daily credit amounts were around 950-1150/day... Now it is 700-850/day... |
KWSN Checklist Send message Joined: 12 Aug 08 Posts: 253 Credit: 275,593,872 RAC: 0
|
After a Google search of "how to determine CPU flop count", some time doing research and testing, this was not a consistent way to judge computer performance. With such a wide variance in calculating FLOPS to award credit for work preformed, a slower computer could get more credit than a faster computer. So the best way would be to make all work units the same size, calculate the flops for the work unit and award credit for this calculation. No matter how fast, or slow, the computation was, credit would be the same. This is the best and most fair solution. I think this is being done by some project already. My two cents worth.
|
|
Send message Joined: 4 Feb 08 Posts: 116 Credit: 17,263,566 RAC: 0 
|
After a Google search of "how to determine CPU flop count", some time doing research and testing, this was not a consistent way to judge computer performance. With such a wide variance in calculating FLOPS to award credit for work preformed, a slower computer could get more credit than a faster computer. So the best way would be to make all work units the same size, calculate the flops for the work unit and award credit for this calculation. No matter how fast, or slow, the computation was, credit would be the same. This is the best and most fair solution. I think this is being done by some project already. My two cents worth. That's exactly what I think too. The projects set a standard credit per standard app running time. They then calculate the credits per unit length. The units carry a standardized credit allotment per unit done. Time in calculation is meaningless. Credit is based on work unit lengths. 
|
Beyond Send message Joined: 15 Jul 08 Posts: 383 Credit: 729,293,740 RAC: 0 
|
For a scholarly discussion of credits for GPU projects and for those who haven't seen it previously: http://www.gpugrid.net/forum_thread.php?id=776&nowrap=true#7556 Makes a lot of sense to me. |
|
The Gas Giant Send message Joined: 24 Dec 07 Posts: 1947 Credit: 240,884,648 RAC: 0
|
The hard part is, some projects are not able to make their wu's the same size due to them either finishing early as they've hit a predetermined cut off, or just due to the fact different wu starting parameters give different wu lengths. |
|
Send message Joined: 26 Jul 08 Posts: 627 Credit: 94,940,203 RAC: 0  
|
After a Google search of "how to determine CPU flop count", some time doing research and testing, this was not a consistent way to judge computer performance. With such a wide variance in calculating FLOPS to award credit for work preformed, a slower computer could get more credit than a faster computer. So the best way would be to make all work units the same size, calculate the flops for the work unit and award credit for this calculation. No matter how fast, or slow, the computation was, credit would be the same. This is the best and most fair solution. I think this is being done by some project already. My two cents worth. If you would have followed the discussions here in the last months you would know that exactly what you propose is done here at Milkyway. I think you confuse it with the benchmark based approach. Actually there are now three projects using the same approach, SETI, GPUGrid and Milkyway. All three projects have counted the amount of necessary operations for a given WU type (determined on server side as it is knows before how much effort is needed to calculate a WU) and this sets the "worth" of a WU. It doesn't matter on which device that WU is calculated, on a fast or a slow CPU or a GPU, it will give the same amount of credit, independently if it needs only a minute on a GPU or two hours on a CPU. So you are a bit too late with your two cents ;) |
|
Send message Joined: 7 Sep 07 Posts: 444 Credit: 5,712,523 RAC: 0 
|
it will give the same amount of credit, independently if it needs only a minute on a GPU or two hours on a CPU. Or nearly two days on my dear old PII 400MHz laptop! :D (About 1 day 19 hours for a credit rate of around 1.2 credits per hour!) |
verstapp Send message Joined: 26 Jan 09 Posts: 589 Credit: 497,834,261 RAC: 0  
|
Perhaps time to upgrade the PII? :) Cheers, PeterV  .
|
TomaszPawel Send message Joined: 9 Nov 08 Posts: 41 Credit: 92,786,635 RAC: 0
|
Hmmmm... Very interesting results.... http://milkyway.cs.rpi.edu/milkyway/results.php?hostid=107881&offset=0&show_names=0&state=3 A proud member of the Polish National Team COME VISIT US at Polish National Team FORUM 
|
|
krahulik Send message Joined: 7 Nov 08 Posts: 14 Credit: 180,768,799 RAC: 0 
|
Hmmmm... You are right. "Very interesting" times for two RV 670... |
|
Send message Joined: 26 Jul 08 Posts: 627 Credit: 94,940,203 RAC: 0
|
Hmmmm... He is running a script that alters the WUs after downloading. He reduces the number of iterations calculated in the WUs resulting in a massive speedup. And I can prove it. I will notify Travis/Anthony of this cheater. In my opinion this guy should be banned. |
|
Beyond Send message Joined: 15 Jul 08 Posts: 383 Credit: 729,293,740 RAC: 0
|
He is running a script that alters the WUs after downloading. He reduces the number of iterations calculated in the WUs resulting in a massive speedup. And I can prove it. I will notify Travis/Anthony of this cheater. In my opinion this guy should be banned. Maybe there could be a validator check to make sure the number of iterations is correct (320?). |
|
TomaszPawel Send message Joined: 9 Nov 08 Posts: 41 Credit: 92,786,635 RAC: 0
|
Hmmmm... If it is real cheat, he should not only be banned but also it is necessary to get his all points back A proud member of the Polish National Team COME VISIT US at Polish National Team FORUM
|
|
Send message Joined: 26 Jul 08 Posts: 627 Credit: 94,940,203 RAC: 0
|
He is running a script that alters the WUs after downloading. He reduces the number of iterations calculated in the WUs resulting in a massive speedup. And I can prove it. I will notify Travis/Anthony of this cheater. In my opinion this guy should be banned. He changed more than this single value. He cuts the computational effort for a WU to about 1/36 of the required one. There are several ways to make such cheating harder with the validator of course an important part of it. But I'm not in the position to decide about the measures to take. |
|
Send message Joined: 26 Jul 08 Posts: 627 Credit: 94,940,203 RAC: 0
|
Hmmmm... I agree. And I think it is proven that he cheats (for a few days already) with full intent and also compromises the results in that course. He "earned" more than 160,000 credits just in the last 6 hours with this cheat. At this rate, he would need about 4 days to arrive at the RAC for the affected machine. That are at least about 2.5 million credits he got through cheating. But in my opinion it wouldn't be enough to remove just the 2.5 million credits. I totally agree with you in that. |
|
TomaszPawel Send message Joined: 9 Nov 08 Posts: 41 Credit: 92,786,635 RAC: 0
|
WTF, i don't know what is going on.... See this..... :( http://milkyway.cs.rpi.edu/milkyway/results.php?hostid=71433&offset=0&show_names=0&state=3 Maby BOINC is not reporting correct times.... or maby not.... A proud member of the Polish National Team COME VISIT US at Polish National Team FORUM
|
|
Send message Joined: 26 Jul 08 Posts: 627 Credit: 94,940,203 RAC: 0
|
WTF, i don't know what is going on.... That are just the CPU times. It's all okay with them. |
©2024 Astroinformatics Group