Viper97
CLASSIFIED ULTRA Member
- Total Posts : 5208
- Reward points : 0
- Joined: 2007/09/07 13:06:18
- Location: Chillin'
- Status: offline
- Ribbons : 11
Re: Major bigadv change (death of the ankle biters)
2013/12/23 12:14:39
(permalink)
TheWolf Yea I started crunching again but didn't make any changes to the rigs from the way they were set. Looking at stats I don't think I'm crunching under the correct project to help with any challenges. Well I did make some changes but don't think I picked the right projects or something. I have no GPU's crunching at the moment only CPU crunching.
As long as you are on the crunching@evga team then you have helped. The Christmas Race is automatic if we join the race. IE all crunching from the time the challenge starts counts towards that challenge if you're on that team.
|
Punchy
CLASSIFIED Member
- Total Posts : 2969
- Reward points : 0
- Joined: 2010/02/06 09:33:05
- Status: offline
- Ribbons : 14
Re: Major bigadv change (death of the ankle biters)
2013/12/23 14:10:57
(permalink)
|
Punchy
CLASSIFIED Member
- Total Posts : 2969
- Reward points : 0
- Joined: 2010/02/06 09:33:05
- Status: offline
- Ribbons : 14
Re: Major bigadv change (death of the ankle biters)
2013/12/23 14:13:55
(permalink)
TheWolf Yea I started crunching again but didn't make any changes to the rigs from the way they were set. Looking at stats I don't think I'm crunching under the correct project to help with any challenges. Well I did make some changes but don't think I picked the right projects or something. I have no GPU's crunching at the moment only CPU crunching.
Your WCG stats show that you haven't returned any work for 139 hours - so maybe you haven't enabled WCG in BOINC manager, and are working on other projects instead.
|
Horvat
CLASSIFIED Member
- Total Posts : 3067
- Reward points : 0
- Joined: 2009/06/16 12:35:24
- Location: Cincinnati, OH
- Status: offline
- Ribbons : 9
Re: Major bigadv change (death of the ankle biters)
2013/12/23 15:16:21
(permalink)
|
cokeman54
SSC Member
- Total Posts : 962
- Reward points : 0
- Joined: 2002/04/02 14:18:49
- Location: Abilene, Texas
- Status: offline
- Ribbons : 3
Re: Major bigadv change (death of the ankle biters)
2013/12/23 17:22:05
(permalink)
For those of you not following FF. PG is getting hammered about core 17 work (lack of it). Todays reply" Top Report this postReply with quoteby VijayPande » Mon Dec 23, 2013 11:44 am We have several new core17 projects on the way as well as more core17 WUs for existing projects. I'll get an update from team members to see where that stands"
|
bcavnaugh
The Crunchinator
- Total Posts : 38516
- Reward points : 0
- Joined: 2012/09/18 17:31:18
- Location: USA Affiliate E5L3CTGE12 Associate 9E88QK5L7811G3H
- Status: offline
- Ribbons : 282
Re: Major bigadv change (death of the ankle biters)
2013/12/23 17:26:09
(permalink)
Yep, but I did get two Core 17's today late today. Today I am back on Core 17 on all my GPUs. 24 Dec 2013
post edited by bcavnaugh - 2013/12/24 08:11:41
|
TheWolf
CLASSIFIED Member
- Total Posts : 3800
- Reward points : 0
- Joined: 2007/11/14 16:05:23
- Location: Moss Point, Ms
- Status: offline
- Ribbons : 9
Re: Major bigadv change (death of the ankle biters)
2013/12/24 06:59:14
(permalink)
Punchy
TheWolf Yea I started crunching again but didn't make any changes to the rigs from the way they were set. Looking at stats I don't think I'm crunching under the correct project to help with any challenges. Well I did make some changes but don't think I picked the right projects or something. I have no GPU's crunching at the moment only CPU crunching.
Your WCG stats show that you haven't returned any work for 139 hours - so maybe you haven't enabled WCG in BOINC manager, and are working on other projects instead.
I know I have PrimeGrid and GPUgrid not sure on WCG but only PrimeGrid is working work at this time.
|
ArtyD42
Superclocked Member
- Total Posts : 212
- Reward points : 0
- Joined: 2013/11/18 18:20:26
- Status: offline
- Ribbons : 0
Re: Major bigadv change (death of the ankle biters)
2013/12/24 16:55:20
(permalink)
bill1024 If you have a better system in mind please go to the FF and give them a clue as what to do to make everyone happy. Please
What needs to happen is for them to stop playing games with the point system. A touchdown in the NFL is 6 points, the point after 1, a field goal 3, a safety 2. It has been this way for many decades. What happens when we change this? Touchdown 2 points. Any kick thru the uprights 1 point. Safety 1 point. Remember that Redskins vs Giants game of 41 to 72? It becomes 20 to 29.
|
ArtyD42
Superclocked Member
- Total Posts : 212
- Reward points : 0
- Joined: 2013/11/18 18:20:26
- Status: offline
- Ribbons : 0
Re: Major bigadv change (death of the ankle biters)
2013/12/24 17:39:29
(permalink)
bill1024 What happens when people all stop folding x wu to fold y wu just to get more points. The points system has never been perfect. So they change the value to move you in the direction they need you to go. Then that causes the values to be off yet again, that makes people complain. Football does not change, but the needs of the researchers, the science, and the hardware does change all the time.
Then set the point after (QRB) on the projects that need it most. x wu and y wu should be equal points via whatever the power to point ratios are supposed to be. Leave the project points alone.
|
ArtyD42
Superclocked Member
- Total Posts : 212
- Reward points : 0
- Joined: 2013/11/18 18:20:26
- Status: offline
- Ribbons : 0
Re: Major bigadv change (death of the ankle biters)
2013/12/24 19:08:10
(permalink)
bill1024 I agree if there are smp units needing to be done they could increase the QRB by a factor of one or two.
Why such a low number? Point grinders will just go with whatever has the QRB number attached to it that is highest which would be related directly to the number of units not yet completed. Like if there's 8 projects out there and we'll just call them A, B, C, D, E, F, G, and H. Each project has 1000 units related to them. A-D need CPU, and E-H need GPU. If your GPU is needing a new work unit and is set in this manner it will look for whichever project is the most important to do work for. If E has 100 units done, F has 200 units done, G has 50 units done, and H has 800 units done it will try to get from Project G. It's got the least work done on it. This would return the highest points due to it not having enough of it completed. It would also lock you into looking for more work from Project G until it's completed. The bonus points for the project would then be sent out based on how much work you completed of it. You want to Quickly Return the project work in order to get the most Bonus points from it. Finished 1/10th of that project's workload? Gain 1/10th that project's points on top of what you already did. If the full project generated 7M points (not counting speedy bonus, that's yours alone) then your bonus would be 700k granted the second it ends for doing 1/10th that project's work. Obviously my number per project is way off the mark and completing the project quickly will give you bonus points based purely on how much time ahead of the deadline you were. But the overall project bonus points will not be given out until the project is complete. Now put the picture in place for this. Our team completes a project's needed folding by 80%. We get 80% of the points, whatever speed return bonus we each compile by beating deadlines, and then 80% of the points on top of that. Of course since it's only given when the entire project is completed we have a reason to focus on and finish a project rather than just go for the biggest point project possible at every second. In short: no completed project = no team QRB bonus. And yes those are team points. If you don't have a team those points vanish. My proposal: End the year, finish the points that people turn in now, and set that as the legacy point system for 2013. Start 2014 with whatever projects are still in progress. Set their project workload to the amount of units left for them to finish. Any new projects get set into the system at 0/X where X is the amount of WU needed to end the project. Set your servers to hand out the highest priority given to the work that is closest to completion. How would this change effect people? Casual users: No change, still fold and do the work units. Team members: Good reason to focus their collective computing power to end projects. "Ankle biters": Still good reason to use the systems without changing anything related to them. Let their processing power be used to complete projects as quickly as they can. If they're focusing on smaller work then they just complete it faster getting more bonus from their return. If they're part of a team then they complete entire projects faster thus granting better QRB team totals. Finally: Make a physical reward. Top 3 teams get their team name etched on a placard or something. Heck even a digital scoreboard wouldn't hurt. Make those top 3 teams a target worth folding for. Like how there's the Stanley's cup. Heck there's even a trophy for League of Legends. Put something like that in here and good god the competition to complete projects would go berserk.
|
Panther-X
New Member
- Total Posts : 54
- Reward points : 0
- Joined: 2013/05/08 07:19:03
- Status: offline
- Ribbons : 1
Re: Major bigadv change (death of the ankle biters)
2013/12/25 04:30:31
(permalink)
ArtyD42...My proposal: End the year, finish the points that people turn in now, and set that as the legacy point system for 2013. Start 2014 with whatever projects are still in progress. Set their project workload to the amount of units left for them to finish. Any new projects get set into the system at 0/X where X is the amount of WU needed to end the project. Set your servers to hand out the highest priority given to the work that is closest to completion. How would this change effect people? Casual users: No change, still fold and do the work units. Team members: Good reason to focus their collective computing power to end projects. "Ankle biters": Still good reason to use the systems without changing anything related to them. Let their processing power be used to complete projects as quickly as they can. If they're focusing on smaller work then they just complete it faster getting more bonus from their return. If they're part of a team then they complete entire projects faster thus granting better QRB team totals. Finally: Make a physical reward. Top 3 teams get their team name etched on a placard or something. Heck even a digital scoreboard wouldn't hurt. Make those top 3 teams a target worth folding for. Like how there's the Stanley's cup. Heck there's even a trophy for League of Legends. Put something like that in here and good god the competition to complete projects would go berserk.
That is a very creative solution and a unique one! Assuming that your idea is implemented, it would require changes at the Server level which would take time and needs to be thoroughly tested. Thus, it just can't be implemented in 2014, rather, it could be tested during 2014 and once confirmed to work, will roll out.
|
ArtyD42
Superclocked Member
- Total Posts : 212
- Reward points : 0
- Joined: 2013/11/18 18:20:26
- Status: offline
- Ribbons : 0
Re: Major bigadv change (death of the ankle biters)
2013/12/25 06:03:52
(permalink)
bill1024 Trouble is the gpu points are based off the smp points as in "equal work for equal pay." So if they raise the smp points they have to raise the gpu points again. Then when those get too high the guys who spend thousands on a 4P system want their points raised. Hence we went full circle. We can go round and round and round. Make it 1 point for 1 wu and the people who really care about the science will build the best system they can afford. Make it like a rank in the Army. or something. A bronze folder and silver folder and a gold, level folder. Or just accept the fact it is not perfect and most likely will never be. Post you Ideas over at FF and see what they think.
Who said anything about changing the base points? My target is the QRB and project completions. I am not suggesting any changes in the GPU and SMP points. Panther-X That is a very creative solution and a unique one! Assuming that your idea is implemented, it would require changes at the Server level which would take time and needs to be thoroughly tested. Thus, it just can't be implemented in 2014, rather, it could be tested during 2014 and once confirmed to work, will roll out.
How hard can it be to set work priority handout by the server to lowest remaining faction workload? I have started folding again btw.
|
Punchy
CLASSIFIED Member
- Total Posts : 2969
- Reward points : 0
- Joined: 2010/02/06 09:33:05
- Status: offline
- Ribbons : 14
Re: Major bigadv change (death of the ankle biters)
2013/12/25 06:24:30
(permalink)
ArtyD42 Any new projects get set into the system at 0/X where X is the amount of WU needed to end the project. Set your servers to hand out the highest priority given to the work that is closest to completion.
It's an interesting idea, but keep in mind that the amount of WU per project is unknown in advance. The later generations of WU are generated based on previous results; some trajectories end quickly and some go on much longer than others. I'm in favor of anything that revamps the completely broken system that exists now.
|
ArtyD42
Superclocked Member
- Total Posts : 212
- Reward points : 0
- Joined: 2013/11/18 18:20:26
- Status: offline
- Ribbons : 0
Re: Major bigadv change (death of the ankle biters)
2013/12/25 08:03:32
(permalink)
Punchy It's an interesting idea, but keep in mind that the amount of WU per project is unknown in advance. The later generations of WU are generated based on previous results; some trajectories end quickly and some go on much longer than others. I'm in favor of anything that revamps the completely broken system that exists now.
Based on the system that exists to track WU's now we have *Project number* (A, B, C) and I have never seen the total digits in each section being greater than 3 digits. 997,002,999 possible. Thru distributed computing I am able to finish a good 30 a month alone on gpu and probably a few extra. I am but one user, and the majority of what I have done is on http://fah-web.stanford.edu/cgi-bin/fahproject.overusingIPswillbebanned?p=8900 Once they've completed this 8900 they can advance to 8901, or 8902, or whatever. Can't be that hard to note all WU's with 8900 that I've completed for team evga due to every one being submitted with a team number. Can't be that hard to note all team numbers out of the 8900s when they finish using them. Can't be that hard to determine how many units were completed, tally that point total up, and reward the teams that did the most work on that project their percentages. Or am I wrong and it's rather impossible to track what I've done for team EVGA? If so why does this exist: http://fah-web2.stanford....e&username=ArtyD42
|
Afterburner
EVGA Forum Moderator
- Total Posts : 17387
- Reward points : 0
- Joined: 2007/09/21 14:41:48
- Location: It's... Classified Yeeeeeeeeeeaaaaaaah........
- Status: offline
- Ribbons : 110
Re: Major bigadv change (death of the ankle biters)
2013/12/25 08:55:02
(permalink)
Ok... Now this is getting interesting!!!
|
ArtyD42
Superclocked Member
- Total Posts : 212
- Reward points : 0
- Joined: 2013/11/18 18:20:26
- Status: offline
- Ribbons : 0
Re: Major bigadv change (death of the ankle biters)
2013/12/25 09:07:27
(permalink)
Afterburner Ok... Now this is getting interesting!!!
Feel free to cross post this over to the folding main forums then. It's not me taking the credit for this either. This is thousands of years of human civilization piled up into one small post. A science known to millions across the globe to one degree or another. One that everybody does even when they don't know they're doing it. Simple math.
|
Punchy
CLASSIFIED Member
- Total Posts : 2969
- Reward points : 0
- Joined: 2010/02/06 09:33:05
- Status: offline
- Ribbons : 14
Re: Major bigadv change (death of the ankle biters)
2013/12/25 09:55:52
(permalink)
ArtyD42
Punchy It's an interesting idea, but keep in mind that the amount of WU per project is unknown in advance. The later generations of WU are generated based on previous results; some trajectories end quickly and some go on much longer than others. I'm in favor of anything that revamps the completely broken system that exists now.
Based on the system that exists to track WU's now we have *Project number* (A, B, C) and I have never seen the total digits in each section being greater than 3 digits. 997,002,999 possible. Thru distributed computing I am able to finish a good 30 a month alone on gpu and probably a few extra. I am but one user, and the majority of what I have done is on http://fah-web.stanford.edu/cgi-bin/fahproject.overusingIPswillbebanned?p=8900 Once they've completed this 8900 they can advance to 8901, or 8902, or whatever. Can't be that hard to note all WU's with 8900 that I've completed for team evga due to every one being submitted with a team number. Can't be that hard to note all team numbers out of the 8900s when they finish using them. Can't be that hard to determine how many units were completed, tally that point total up, and reward the teams that did the most work on that project their percentages. Or am I wrong and it's rather impossible to track what I've done for team EVGA? If so why does this exist: http://fah-web2.stanford....e&username=ArtyD42
We seem to be talking on different wavelengths. In your earlier message you said the servers could prioritize WU by the number remaining in a project - or maybe I misunderstood. My point is that the number of WU in a project is not known until the project is complete, so you can't prioritize that way. Some runs are done, then based on those results, the best ones are picked to continue and others are abandoned. At least that's my understanding - Panther-X might be able to elaborate.
|
ArtyD42
Superclocked Member
- Total Posts : 212
- Reward points : 0
- Joined: 2013/11/18 18:20:26
- Status: offline
- Ribbons : 0
Re: Major bigadv change (death of the ankle biters)
2013/12/25 10:01:49
(permalink)
Editing as I go and unfortunately will be away from keyboard soon... bill1024 If you're going to tie it to team work, what about the small teams with only 10, 20 folders? What about the folders like me that fold for 2 or more teams?
http://fah-web2.stanford.edu/cgi-bin/main.py?qtype=userpage&username=bill1024Correct? I see three teams, each WU is tracked per team. Punchy We seem to be talking on different wavelengths. In your earlier message you said the servers could prioritize WU by the number remaining in a project - or maybe I misunderstood. My point is that the number of WU in a project is not known until the project is complete, so you can't prioritize that way. Some runs are done, then based on those results, the best ones are picked to continue and others are abandoned. At least that's my understanding - Panther-X might be able to elaborate. A WU in a project is not known. A set maximum is. Lets start with something that is known. 100,000 work units. I'm certain more than this exists. Heck team evga has finished [font="'pt sans', sans-serif, sans-serif"]18,079,435 WU. If we all piled into the same project number then we'd have 180 work projects completed. Obviously that is not true at this time. Getting it to function would take some fiddling with setting a bar. Maybe it needs to be 500K WU, maybe 1M, and I really don't have the data for total units completely folded to tell you an exact number. Somewhere in this mess it exists: http://fah-web.stanford.edu/psummary.html bill1024 As of right now the most complaints are with the points for SMP. Lets take a 4P 48 core. Make it so smp can get close to what it can get with bigadv. One way to do it is improve the QRB for the fastest most powerful . Say the qrb for a 36 hour return is 3x base points. Then make it 4x for 24, 5x for 18, 6x for 12, 7x ....... Up to say 15x for 2 hour return. Up to the point is close to being the same. Of courst the numbers have to be worked. Right now it is just 3x fixed, and you get extra PPD by doing more in a day w/faster system. I say make the bonus multiplier go up as the return time goes down,
You're talking about deleting bigadv folders there. I'm fine with them doing the high end work units as they currently have been. They function. No need to mess with them unless they stop doing so. Their point totals don't have any need to be changed. So if a 4P system completes the work within the deadline it's fine. If a 2P system completes that same work within the deadline it's fine. There's absolutely no point in saying that a 1P system should = a 4P system. That is cooking with napalm. A 4P system should have projects requiring 4 times the work within the same time as a 1P system specifically because it has 4 times the processors. They also should get 4 times the points of a 1P system running the same time. That's the benefit of running higher end machines. They have more that they can do with them in less time. Trying to **** around with these people is not only poor judgement but bad business. And how many people stay when bad business happens to them from a store? Stanford is looking for our time and computer processes. If they want a steady stream of customers then they need to do right by those customers. Otherwise they'll take their business elsewhere. I can just as easily switch back to litecoin mining or crunching. If they want me to fold then they need to get my interest back in folding. I'm folding right now, and they can thank evga for that. But once I've capped my evga goal out what point do I have to keep going? I won't get any benefit for doing more towards my specified goal. I won't get any recognition outside of the fact that I've done what I've done and it can be seen on that digital paper. It'll advance research... maybe. And my goal is to help charity. I don't think Curecoin will do that at this time. Bitcoin would be out of reach for my system. Litecoin I can get to work towards raising money for charity by donating my time and energy from my computer. What helps me reach my goal by folding? The answer is not much. 10 evga bucks a month for doing 500k points. In a few years I'll be able to invest that directly into getting something for a charity. I'll keep doing that. But once I hit that 500k point total my benefit from folding to my goal of helping charity is gone. I've pretty much figured out that the evga data on it completes at 9pm central standard time. So when I'm about to cross that 500k goal I can set the work units to finish and dedicate my processing power elsewhere. I'm not really in it for the points as others are. I probably can do better without folding on getting more towards charity. Folding trumps these only so long as there is a secondary benefit to folding. That's why I support team EVGA. The people here have been good to me and helped me figure out how to do better with what hardware I have. I will always fold that 500k a month. I'm not the person looking at those points with the idea that they're NBA professionals getting those points. I'm also not trying to squash their points for my benefit. To be as blunt as possible: http://fah-web.stanford.edu/cgi-bin/main.py?qtype=teamstats You want this to mean something? Then it needs to be MADE to mean something. This is folding's greatest tragedy at this time. These points don't mean anything. That needs to change before they start alienating everybody who's been putting in the bigadv work. So before anyone cares about the smp, or the qrb, or the team stats there needs to be something recognizing those people that isn't just a crappy digital paper that can be forged. Yes... forged. Those certificates can be edited. A physical object (even one that I can never reach personally) can't. And if 100 bucks is enough to make one exist I'll donate that myself. Do right by your community of folders if you want to keep them. This current method of throwing us all on the back burner because none of us matter will kill folding in the long run.
post edited by ArtyD42 - 2013/12/25 10:57:25
|
Punchy
CLASSIFIED Member
- Total Posts : 2969
- Reward points : 0
- Joined: 2010/02/06 09:33:05
- Status: offline
- Ribbons : 14
Re: Major bigadv change (death of the ankle biters)
2013/12/25 10:28:50
(permalink)
bill1024 I say make the bonus multiplier go up as the return time goes down,
That's how it works already.
|
ArtyD42
Superclocked Member
- Total Posts : 212
- Reward points : 0
- Joined: 2013/11/18 18:20:26
- Status: offline
- Ribbons : 0
Re: Major bigadv change (death of the ankle biters)
2013/12/25 11:02:57
(permalink)
ok, you guys kept going while I made an edit to my post trying to respond to everything said to me before. I finished my edit. I have to go to relatives now. Be back sometime tonightish.
|
Viper97
CLASSIFIED ULTRA Member
- Total Posts : 5208
- Reward points : 0
- Joined: 2007/09/07 13:06:18
- Location: Chillin'
- Status: offline
- Ribbons : 11
Re: Major bigadv change (death of the ankle biters)
2013/12/25 12:13:56
(permalink)
Wow.. I have folded for three teams. EVGA, [H] and another.
|
ArtyD42
Superclocked Member
- Total Posts : 212
- Reward points : 0
- Joined: 2013/11/18 18:20:26
- Status: offline
- Ribbons : 0
Re: Major bigadv change (death of the ankle biters)
2013/12/25 12:54:40
(permalink)
bill1024 It is better to make a new post than to make a huge edit to your old post; as some may never see it if they keep looking forward. We can go back and forth all day long and will not make any difference. Go here and have your voice heard. https://foldingforum.org/ I am still not sure why you posted my FAH stats? What's your point?
Was actually your point. The work units are tracked by team number. Whatever work you do for the wolves or evga or abit wolves is attributed to those teams respectively. That's what you did with the team number in the first place.
|
Panther-X
New Member
- Total Posts : 54
- Reward points : 0
- Joined: 2013/05/08 07:19:03
- Status: offline
- Ribbons : 1
Re: Major bigadv change (death of the ankle biters)
2013/12/25 13:29:00
(permalink)
Please note that I don't have any additional knowledge of how the Servers work than someone who know hows how to read the Server Status page. AFAIK, only a handful of people know in detail how the Server system works and they tend to be either Pande Group Members or the Developers themselves. All of my information regarding the Server Status page and the workings of F@H Servers can be view here ( post) and my understanding is based on observations and donor reports. Thus, if you think that I have made a mistake (quite possible), would be glad to hear it. ArtyD42...How hard can it be to set work priority handout by the server to lowest remaining faction workload?... Regarding the priority system, changing the "number" seems easy. However, there are different versions of the Work Servers so it is possible that not all systems have the same method. The researcher who manages the server will have to continuously monitor the system and adjust the priority. That would cut into the researcher's time and if you tweak something too much, an accident may happen which would have a negative effect on a huge number of systems. Moreover, not all SMP WUs or GPU WUs are present on the same Server. They are across different servers and owners. Thus, it would require a coordinated effort across all the researchers to change the priority. I honestly don't think that it is feasible with the current infrastructure. However, if a new system is developed which dynamically adjusts the priority of WUs with a possibility of a manual override, it would be interesting to see how it pans out. ArtyD42...Based on the system that exists to track WU's now we have *Project number* (A, B, C) and I have never seen the total digits in each section being greater than 3 digits. 997,002,999 possible.... The highest Generation value that I have seen on the Official Folding forum was something above 1000 (a very quick search revels Gen 2247 for FahCore_11 WU). However, please note that not all "branches" are successful as we have bad WUs, usually less than 1% for released Projects. ArtyD42...Can't be that hard to note all WU's with 8900 that I've completed for team evga due to every one being submitted with a team number. Can't be that hard to note all team numbers out of the 8900s when they finish using them. Can't be that hard to determine how many units were completed, tally that point total up, and reward the teams that did the most work on that project their percentages. Or am I wrong and it's rather impossible to track what I've done for team EVGA?... You are correct if we only focus on a single Project that finished. However, PSummary (Projects available to all donors without any client-type set) lists 527 Projects currently ready to be assigned. And that's a lot of data to be processed. AFAIK, during the early days, donors could get a breakdown of what Projects they contributed to and a lot more details. Unfortunately, as donors increased, the volume of information increased and in order for the Stats system to function without issues, these "extra" bits were no longer processed. I believe that they still collect the raw data but don't process it. Maybe, in a future they gain sufficient computer power to process the entire data every hour for 236K+ donors. Alternatively, they could do it on a yearly basis if they have sufficient resources. Punchy...My point is that the number of WU in a project is not known until the project is complete, so you can't prioritize that way. Some runs are done, then based on those results, the best ones are picked to continue and others are abandoned. At least that's my understanding - Panther-X might be able to elaborate. You are correct that we don't know the end point, but we know the start point, i.e. a researcher will set the server to generate 100 WUs (just for example) and it can take a couple of hours (that is what we faced during the testing of FahCore_17 over the IRC) to finish. Once done, the WUs are then assigned to the Clients. Sometimes, there is a lower limit of WUs and if that limit is reached, the server stops assigning WUs. Assuming that the Server assigned 80 WUs, once they are returned back, the next gen is generated and ready to be assigned. This continues until: 1) We have an unstable trajectory (bad WU) 2) The researcher decides that it is now time to check the returned data and could either lower the priority or temporarily suspend those WUs. Sometime, the data generated from these "initial" runs can provide a valuable insight on how to approach this issue and could potentially tweak the system or even release another set of Projects with some minor differences in the set-up (temperature, velocity, etc). Now, what got me by surprise was the announcement of a new Client Type along with a back-end Server (Yep, I wasn't aware of this until I read Dr. Vijay's Blog). This would mean that V7 will still be developed and supported and there will be a new type of Client. With the mention of a new back-end Server system, my guess is that maybe the point system could be different, maybe it would be F@H Rebooted. I don't know as of now and thus, am eagerly awaiting for more details. in the end, the future isn't set in stone, thus, changes can always happen. I hope that it happens for the betterment of the project. Kudos to everyone who helps out doing charity work to help improve humanity and paint a brighter picture of the future, be it by folding, crunching, donating, etc. They each have their own unique place amongst the contributors and I appreciate all of them
post edited by Panther-X - 2013/12/25 13:31:28
|
clo007
CLASSIFIED Member
- Total Posts : 3193
- Reward points : 0
- Joined: 2006/02/09 11:20:25
- Status: offline
- Ribbons : 11
Re: Major bigadv change (death of the ankle biters)
2013/12/26 07:33:07
(permalink)
|
clo007
CLASSIFIED Member
- Total Posts : 3193
- Reward points : 0
- Joined: 2006/02/09 11:20:25
- Status: offline
- Ribbons : 11
Re: Major bigadv change (death of the ankle biters)
2013/12/26 07:37:59
(permalink)
ArtyD42
bill1024 I agree if there are smp units needing to be done they could increase the QRB by a factor of one or two.
Why such a low number? Point grinders will just go with whatever has the QRB number attached to it that is highest which would be related directly to the number of units not yet completed. Like if there's 8 projects out there and we'll just call them A, B, C, D, E, F, G, and H. Each project has 1000 units related to them. A-D need CPU, and E-H need GPU. If your GPU is needing a new work unit and is set in this manner it will look for whichever project is the most important to do work for. If E has 100 units done, F has 200 units done, G has 50 units done, and H has 800 units done it will try to get from Project G. It's got the least work done on it. This would return the highest points due to it not having enough of it completed. It would also lock you into looking for more work from Project G until it's completed. The bonus points for the project would then be sent out based on how much work you completed of it. You want to Quickly Return the project work in order to get the most Bonus points from it. Finished 1/10th of that project's workload? Gain 1/10th that project's points on top of what you already did. If the full project generated 7M points (not counting speedy bonus, that's yours alone) then your bonus would be 700k granted the second it ends for doing 1/10th that project's work. Obviously my number per project is way off the mark and completing the project quickly will give you bonus points based purely on how much time ahead of the deadline you were. But the overall project bonus points will not be given out until the project is complete. Now put the picture in place for this. Our team completes a project's needed folding by 80%. We get 80% of the points, whatever speed return bonus we each compile by beating deadlines, and then 80% of the points on top of that. Of course since it's only given when the entire project is completed we have a reason to focus on and finish a project rather than just go for the biggest point project possible at every second. In short: no completed project = no team QRB bonus. And yes those are team points. If you don't have a team those points vanish. My proposal: End the year, finish the points that people turn in now, and set that as the legacy point system for 2013. Start 2014 with whatever projects are still in progress. Set their project workload to the amount of units left for them to finish. Any new projects get set into the system at 0/X where X is the amount of WU needed to end the project. Set your servers to hand out the highest priority given to the work that is closest to completion. How would this change effect people? Casual users: No change, still fold and do the work units. Team members: Good reason to focus their collective computing power to end projects. "Ankle biters": Still good reason to use the systems without changing anything related to them. Let their processing power be used to complete projects as quickly as they can. If they're focusing on smaller work then they just complete it faster getting more bonus from their return. If they're part of a team then they complete entire projects faster thus granting better QRB team totals. Finally: Make a physical reward. Top 3 teams get their team name etched on a placard or something. Heck even a digital scoreboard wouldn't hurt. Make those top 3 teams a target worth folding for. Like how there's the Stanley's cup. Heck there's even a trophy for League of Legends. Put something like that in here and good god the competition to complete projects would go berserk.
I will ask my buddy if he will make a traveling trophy for this. He has made the official tournament trophies for Quakecon since 2009. http://www.alteredalloy.com/portfolio/portfolio.html These are some of the trophies he has made. He bought a laser cutter a couple of months back and can pretty much make anything now. Here is his esty store: http://www.etsy.com/shop/alteredalloy I am not promoting him, just take a look at what he could do for us. If you guys are serious, let me know and I will talk to him.
post edited by clo007 - 2013/12/26 07:40:33
|
kerryd
FTW Member
- Total Posts : 1014
- Reward points : 0
- Joined: 2007/02/28 18:14:32
- Location: Wash. Affiliate Code: TGVNCLF86N
- Status: offline
- Ribbons : 0
Re: Major bigadv change (death of the ankle biters)
2013/12/26 20:54:21
(permalink)
I call BULL CRAP on "equal work for equal pay.". If it was "equal work for equal pay." you would not lose 50% or more ppd going from bigadv to smp It should be "equal work for equal pay." for smp gpu AND bigadv if they keep saying thats what we are getting. I do not care they are going to piss some one off no matter what they do but most ppl do fold smp/gpu
post edited by kerryd - 2013/12/26 20:56:02
|
Grandpa_01
New Member
- Total Posts : 91
- Reward points : 0
- Joined: 2012/04/28 20:59:00
- Status: offline
- Ribbons : 0
Re: Major bigadv change (death of the ankle biters)
2013/12/26 22:31:41
(permalink)
kerryd I call BULL CRAP on "equal work for equal pay.". If it was "equal work for equal pay." you would not lose 50% or more ppd going from bigadv to smp It should be "equal work for equal pay." for smp gpu AND bigadv if they keep saying thats what we are getting. I do not care they are going to piss some one off no matter what they do but most ppl do fold smp/gpu
It is not = work for = pay, it is = pay for = science. if it was = work for = pay then smp would have a higer value than GPU the largest GPU WU is 40,000 atoms the largest SMP is 167,000 atoms the bigadv are 1,340,422 atoms now you can figure out the wattage and investment it takes to complete each if you wish but I am pretty sure you do not want a = work for = pay scenario because GPU would go way down then there would be a whole other group of really pissed off people. No matter how you look at it smp needs to go up in value in order to attract people to them, Bigadv qualifications are going to change whether we want them to or not Bigadv was designed for top end machines and as the top end moves so will the requirements. They need a better roadmap and hopefully that will happen but that still is not going to change the fact that the requirements are going to continue to go up.
|
Xavier Zepherious
CLASSIFIED ULTRA Member
- Total Posts : 6746
- Reward points : 0
- Joined: 2010/07/04 12:53:39
- Location: Medicine Hat ,Alberta, Canada
- Status: offline
- Ribbons : 16
Re: Major bigadv change (death of the ankle biters)
2013/12/26 22:44:41
(permalink)
equal work for equal pay refers to SMP work on GPU - since core 17 can now do this - and why GPU points are so high (because a GPU finishes them quicker) so if SMP goes up GPU points have to go up I have no issue with this and nor should bigadv users - you can always opt to go to GPU farms too (switch or do both) Bigadv was not included in this category - It stands as a special subset of high end exploratory subset Biologist and Stanford wish to push on testing how far they can push their science and technology (work that can't be performed on GPU) however dropping from BIGADV on hardware to SMP shouldn't mean getting 1/4 -1/3 the points Id be happy at 50-70% of bigadv
post edited by Xavier Zepherious - 2013/12/26 22:46:18
|
BadBertie
SSC Member
- Total Posts : 852
- Reward points : 0
- Joined: 2012/08/10 17:09:48
- Location: France
- Status: offline
- Ribbons : 1
Re: Major bigadv change (death of the ankle biters)
2013/12/26 22:56:04
(permalink)
I'm still working my way through the comments here, so sorry if I repeat previous posts, but would it not be possible to link a second 24 core board from another to make a 48 core machine? I seem to recall reading something along those lines when I was investigating folding 12 months ago. Whether you can do it with all Supermicro/Ubuntu boards/OS with dual Lan is another matter. Did anyone look at that possibility at all?
|
Grandpa_01
New Member
- Total Posts : 91
- Reward points : 0
- Joined: 2012/04/28 20:59:00
- Status: offline
- Ribbons : 0
Re: Major bigadv change (death of the ankle biters)
2013/12/26 23:06:41
(permalink)
A explanation from VJ on = pay for = science not work but science = pay for = work means it takes x amount of energy to do x amount of work = y pay the do not have a = work for = pay points system. https://foldingforum.org/...1&p=253960#p253960
|