Bigadv points change
Moderators: Site Moderators, FAHC Science Team
Re: Bigadv points change
I'm sorry but I simply can't accept that GPUs and CPUs shouldn't be compared. They cost money to procure, they cost money to run and they produce points, which supposedly relate to the scientific value of the work. It doesn't make much sense not to compare the scientific productivity per $ and, more importantly in the long term, the points (scientific value)/KWh. In making the comparison, I found that even with the reduction in ppd on bigadv, it is still far more cost effective to spend folding dollars on bigadv rigs, to the exclusion of the other forms. That is what I perceive to be the real problem. While I haven't gone so far as to move or turn off clients, every $ I've invested in new equipment has been to fold bigadv and I'm far from the only one doing that, judging from this thread.
-
- Posts: 76
- Joined: Tue Apr 29, 2008 11:02 pm
- Hardware configuration: XP-32 Pro SP-3
Antec NSK-2480 with two Thermaltake 120mm Smart Fans
Gigabyte ga-ma78gm-s2h 780G IGP
BE-2350 with 10.5 x multiplier, 1.250V in BIOS, clock at 272 (2.856GHz)
EVGA 8800 GS
Ninja Mini CPU HS
GeIL 4GB (2 x 2GB) 240-Pin DDR2 SDRAM DDR2 800
Seagate 500GB SATA hard drive
ASUS 18X DVD±R DVD Burner PATA Model DRW-1814BL
Re: Bigadv points change
Please to be seeing someone else's post about forming classes within FAH so that points on GPU aren't the same as points on bigadv, smp or uni.Grandpa_01 wrote:GPU's really should not be compared to CPU's totally different type of work and value of work. From what I understand in laymen’s terms, GPU's eliminate paths and find paths to follow. CPU do the intense work of blazing the trail. I could be wrong but that is the way I understand it.
Not that I'm taking a position on that suggestion.
-
- Posts: 1122
- Joined: Wed Mar 04, 2009 7:36 am
- Hardware configuration: 3 - Supermicro H8QGi-F AMD MC 6174=144 cores 2.5Ghz, 96GB G.Skill DDR3 1333Mhz Ubuntu 10.10
2 - Asus P6X58D-E i7 980X 4.4Ghz 6GB DDR3 2000 A-Data 64GB SSD Ubuntu 10.10
1 - Asus Rampage Gene III 17 970 4.3Ghz DDR3 2000 2-500GB Segate 7200.11 0-Raid Ubuntu 10.10
1 - Asus G73JH Laptop i7 740QM 1.86Ghz ATI 5870M
Re: Bigadv points change
I am assuming that you originally invested in the GPU's because that is where the Technology of the day was and the had the best point return for the $$$ spent and you most likely made a pile of points from them. Well apparently the Technology has devolved better for protein folding in the CPU end than the GPU end of things and for the time being the GPU hay-day is over and it is now time for the CPU. It really does not make any sense to me to just give more pay to something doing less work.ChasR wrote:I'm sorry but I simply can't accept that GPUs and CPUs shouldn't be compared. They cost money to procure, they cost money to run and they produce points, which supposedly relate to the scientific value of the work. It doesn't make much sense not to compare the scientific productivity per $ and, more importantly in the long term, the points (scientific value)/KWh. In making the comparison, I found that even with the reduction in ppd on bigadv, it is still far more cost effective to spend folding dollars on bigadv rigs, to the exclusion of the other forms. That is what I perceive to be the real problem. While I haven't gone so far as to move or turn off clients, every $ I've invested in new equipment has been to fold bigadv and I'm far from the only one doing that, judging from this thread.
I myself used to fold with mostly GPU's but realised that the CPU's were doing more work and made more points per watt so as I upgraded I switched to bigadv machines which apparently you are doing at this time. Stanford has set the bar where they want it and that bar says bigadv right now. If they come out with a GPU WU that is more valuable to them in some way then I say raise the value to reflect the work. But do not raise the value just to please some people. They basicly already did that when they dropped the value of the bigadv.
I do not think it would be very wise on Stanford part to add insult to injury, there is already enough grumbling going on in other forums about conspiracy theory's and going to different DC projects. I have a feeling there may be a bit of an exodus if they raise the value of GPU folding for no reason other than people griping.
2 - SM H8QGi-F AMD 6xxx=112 cores @ 3.2 & 3.9Ghz
5 - SM X9QRI-f+ Intel 4650 = 320 cores @ 3.15Ghz
2 - I7 980X 4.4Ghz 2-GTX680
1 - 2700k 4.4Ghz GTX680
Total = 464 cores folding
Re: Bigadv points change
If GPU's get a point increase, after bigadv just got whacked, I suspect the science output will take a hammering as disgruntled folders take their CPU&GPUs and go play another game.Grandpa_01 wrote:ChasR wrote:I do not think it would be very wise on Stanford part to add insult to injury, there is already enough grumbling going on in other forums about conspiracy theory's and going to different DC projects. I have a feeling there may be a bit of an exodus if they raise the value of GPU folding for no reason other than people griping.
H.
Re: Bigadv points change
@ Grandpa
I did add a lot of of GPUs and I would agree with you if it were truly the march of technology that brought us to this point. It simply isn't the root cause. It is the points scheme, the unsustainable exponential QRB and all the bigadv WUs after p2684. I've said this before, but once again, a 2600K (4.8 GHz) is about 12x as powerful as a P4 640 (3.8 GHz), but was awarded 400x the ppd prior to the reduction, comparing the uniprocessor to bigadv client. The same 2600K is about 2.3x as powerful as a q6600 (3.4 GHz) but was awarded 7.5 times the ppd comparing the smp client to the -bigadv client. A change was needed and a change was made. It's a bit closer now and I applaud Pande Group for having the guts to make unpleasant corrections.
I did add a lot of of GPUs and I would agree with you if it were truly the march of technology that brought us to this point. It simply isn't the root cause. It is the points scheme, the unsustainable exponential QRB and all the bigadv WUs after p2684. I've said this before, but once again, a 2600K (4.8 GHz) is about 12x as powerful as a P4 640 (3.8 GHz), but was awarded 400x the ppd prior to the reduction, comparing the uniprocessor to bigadv client. The same 2600K is about 2.3x as powerful as a q6600 (3.4 GHz) but was awarded 7.5 times the ppd comparing the smp client to the -bigadv client. A change was needed and a change was made. It's a bit closer now and I applaud Pande Group for having the guts to make unpleasant corrections.
-
- Posts: 1122
- Joined: Wed Mar 04, 2009 7:36 am
- Hardware configuration: 3 - Supermicro H8QGi-F AMD MC 6174=144 cores 2.5Ghz, 96GB G.Skill DDR3 1333Mhz Ubuntu 10.10
2 - Asus P6X58D-E i7 980X 4.4Ghz 6GB DDR3 2000 A-Data 64GB SSD Ubuntu 10.10
1 - Asus Rampage Gene III 17 970 4.3Ghz DDR3 2000 2-500GB Segate 7200.11 0-Raid Ubuntu 10.10
1 - Asus G73JH Laptop i7 740QM 1.86Ghz ATI 5870M
Re: Bigadv points change
I can not argue with your numbers because I have no idea if they are correct or not. But the numbers are irrelevant when it comes to the QRB or Quick Return Bonus which is what it means. And Stanford assigned a value to the complexity the amount of science done and the amount of time it took to get the science back and in their view that is what it is worth to them. What you or I want is really irrelevant as far as scientific value goes. What brought us to this point is Stanford said hey this work is worth this much to us, do it and we will give you X. I am pretty sure Stanford sets the values to pretty much be in line with what they want and need at the time to advance the science.
Technology changes and there needs change the project needs X to improve and move forward. If you want more of X then give them Y and you will get X. I still say if Stanford comes up with a GPU WU that is worth more then give the more. But don't just give more because people are griping. What good will that do the project ? or will it hurt the project by costing folders?
And by the way I have always felt the point system was lopsided whether it was ATI vs nVidia or GPU vs CPU in the GPU hay-day or the current CPU vs GPU. There has always been to big of a spread 1 way or the other. And were the GPU guys griping in the GPU hay-day. I can answer that question with 100% confidence (No)
Technology changes and there needs change the project needs X to improve and move forward. If you want more of X then give them Y and you will get X. I still say if Stanford comes up with a GPU WU that is worth more then give the more. But don't just give more because people are griping. What good will that do the project ? or will it hurt the project by costing folders?
And by the way I have always felt the point system was lopsided whether it was ATI vs nVidia or GPU vs CPU in the GPU hay-day or the current CPU vs GPU. There has always been to big of a spread 1 way or the other. And were the GPU guys griping in the GPU hay-day. I can answer that question with 100% confidence (No)
2 - SM H8QGi-F AMD 6xxx=112 cores @ 3.2 & 3.9Ghz
5 - SM X9QRI-f+ Intel 4650 = 320 cores @ 3.15Ghz
2 - I7 980X 4.4Ghz 2-GTX680
1 - 2700k 4.4Ghz GTX680
Total = 464 cores folding
Re: Bigadv points change
It will always be a balancing act to try and keep ppd equal across different projects running such different hardware,
but the Gpu has it's limitations as far as a QRB or other bonus idea's are concerned, we only need to look back as far
as the P6811 Gpu Wu's which were the first of what could be termed "Big", most ran them ok, but a significant number
had Wu failures and overheating cards even at stock settings, while some would argue that they should've been continued
for those with Gpu's that could produce good points from them, overclocked and or watercooled, but it wouldn't be a good idea from a science viewpoint to have a project open to failed Wu's dependant on donor hardware, For a current or future project that demands big Wu's and fast return times Cpu smp systems are for now still the best option, but the most expensive one for donors who rightly or wrongly expect an incentive to build such machines, And I'm not talking of heavily overclocked quadcore Cpu rig's that can gain bonus points by being inside a preffered deadline, but those (again, I'm not one of them) with expensive, multi Cpu,dedicated systems who can return Big Wu's quickly with both ease and stability, and if something like a reduction in deadlines turns those who invest so much in the project into some sort of "Folding Elite" then they should be both admired and applauded for their efforts, And that is just now, if in the future a Gpu client is released that will allow 2, 3 or even 4 way SLi or crossfire with Gpu Bigadv Wu's to suit then those that build rigs to suit those demands should also be rewarded accordingly, with any bonus not set just by their speed compared to a single card, but by how crucial it is to process the Wu and how fast it is deemed to be needed at the time, Having a few so called "Elite" machines or donors doesn't in any way devalue the contribution of other donors, it enhances it by speeding up the science of the project as a whole and as a global team effort.
but the Gpu has it's limitations as far as a QRB or other bonus idea's are concerned, we only need to look back as far
as the P6811 Gpu Wu's which were the first of what could be termed "Big", most ran them ok, but a significant number
had Wu failures and overheating cards even at stock settings, while some would argue that they should've been continued
for those with Gpu's that could produce good points from them, overclocked and or watercooled, but it wouldn't be a good idea from a science viewpoint to have a project open to failed Wu's dependant on donor hardware, For a current or future project that demands big Wu's and fast return times Cpu smp systems are for now still the best option, but the most expensive one for donors who rightly or wrongly expect an incentive to build such machines, And I'm not talking of heavily overclocked quadcore Cpu rig's that can gain bonus points by being inside a preffered deadline, but those (again, I'm not one of them) with expensive, multi Cpu,dedicated systems who can return Big Wu's quickly with both ease and stability, and if something like a reduction in deadlines turns those who invest so much in the project into some sort of "Folding Elite" then they should be both admired and applauded for their efforts, And that is just now, if in the future a Gpu client is released that will allow 2, 3 or even 4 way SLi or crossfire with Gpu Bigadv Wu's to suit then those that build rigs to suit those demands should also be rewarded accordingly, with any bonus not set just by their speed compared to a single card, but by how crucial it is to process the Wu and how fast it is deemed to be needed at the time, Having a few so called "Elite" machines or donors doesn't in any way devalue the contribution of other donors, it enhances it by speeding up the science of the project as a whole and as a global team effort.
Re: Bigadv points change
There's a saying, "what if you gave a parade and nobody came." The "elitist" label and attitude doesn't win friends or supporters in this endeavor and is rather disheartening to read to say the least. This isn't about GPU versus SMP. This is about the huge disparity within the SMP QRB for the very large units that is out of proportion to the other not so large SMP "as perceived by the general donor contributing population." What it is doing is causing folks to question whether it is worth their time and money to continue contributing for the other areas of the PG project. If the big systems contribute so much to the project, then why waste the resources of the other systems? Well one big reason for the small stuff is that if there wasn't the groundswell of public opinion created through all the "little people" to the project, then it's quite likely that the big organizations contributing the big money and support to the PG effort wouldn't be there to contribute it. So it is a real double edged sword for PG to manage the public relations aspect of this issue. Otherwise they wouldn't even be wasting their time on it. No body likes to have their face rubbed in it. But that is what is happening here, and people are taking notice. There isn't really any "team competition" with this project anymore as the whole point system is just totally out of whack. Yes, the big boys/girls contribute a lot of work here now, and it's all basically a game between them. Other folks come and go. Look at all the large teams and their membership rosters, the majority of their members don't contribute. On their forums a good number of the folks have moved on to other projects where there is more interest and competition among members and teams. This points issue is just accelerating that migration. But then, who cares............. the big machines will solve it all.
Re: Bigadv points change
I Fold for the EVGA Team and every time I've checked, the majority of our Team Points come from Folders that work the standard WU's (CPU and GPU). That is a big reason that our team is strong and I like that very much. No doubt, the people with larger, more powerful rigs are making a large contribution too, but we don't "hang our hat" on just one size of Folding contributor.
People do come and go on our team for a variety of reasons, but seldom is that reason due to Pande's Folding Points structure. When I first began Folding, I was running just Uni-Processor WU's at a snail's pace. I was very glad to do it and never begrudged anyone that had larger, more powerful rigs. Their Folding rigs were very interesting to me and more of an incentive to move forward for me into a larger investment in Folding equipment. I think for most Folders, we all share a very common purpose in this work and realize that we all are accomplishing the tasks no matter what rigs we choose to use.
People do come and go on our team for a variety of reasons, but seldom is that reason due to Pande's Folding Points structure. When I first began Folding, I was running just Uni-Processor WU's at a snail's pace. I was very glad to do it and never begrudged anyone that had larger, more powerful rigs. Their Folding rigs were very interesting to me and more of an incentive to move forward for me into a larger investment in Folding equipment. I think for most Folders, we all share a very common purpose in this work and realize that we all are accomplishing the tasks no matter what rigs we choose to use.
-
- Posts: 1122
- Joined: Wed Mar 04, 2009 7:36 am
- Hardware configuration: 3 - Supermicro H8QGi-F AMD MC 6174=144 cores 2.5Ghz, 96GB G.Skill DDR3 1333Mhz Ubuntu 10.10
2 - Asus P6X58D-E i7 980X 4.4Ghz 6GB DDR3 2000 A-Data 64GB SSD Ubuntu 10.10
1 - Asus Rampage Gene III 17 970 4.3Ghz DDR3 2000 2-500GB Segate 7200.11 0-Raid Ubuntu 10.10
1 - Asus G73JH Laptop i7 740QM 1.86Ghz ATI 5870M
Re: Bigadv points change
RenixRenix wrote:There's a saying, "what if you gave a parade and nobody came." The "elitist" label and attitude doesn't win friends or supporters in this endeavor and is rather disheartening to read to say the least. This isn't about GPU versus SMP. This is about the huge disparity within the SMP QRB for the very large units that is out of proportion to the other not so large SMP "as perceived by the general donor contributing population." What it is doing is causing folks to question whether it is worth their time and money to continue contributing for the other areas of the PG project. If the big systems contribute so much to the project, then why waste the resources of the other systems? Well one big reason for the small stuff is that if there wasn't the groundswell of public opinion created through all the "little people" to the project, then it's quite likely that the big organizations contributing the big money and support to the PG effort wouldn't be there to contribute it. So it is a real double edged sword for PG to manage the public relations aspect of this issue. Otherwise they wouldn't even be wasting their time on it. No body likes to have their face rubbed in it. But that is what is happening here, and people are taking notice. There isn't really any "team competition" with this project anymore as the whole point system is just totally out of whack. Yes, the big boys/girls contribute a lot of work here now, and it's all basically a game between them. Other folks come and go. Look at all the large teams and their membership rosters, the majority of their members don't contribute. On their forums a good number of the folks have moved on to other projects where there is more interest and competition among members and teams. This points issue is just accelerating that migration. But then, who cares............. the big machines will solve it all.
I participate and read allot of folding forums and I simply find very little of what you are saying actually taking place. From what I have seen there is actually a very small minority of people griping about the point system. (although a very vocal minority) most basically say what it is , is what it is. Who is rubbing you face in what.? People have come and gone from the beginning of FAH before there was any point discrepancies but guess what fah is still growing not shrinking the last I read said FAH computation power is greater than the world’s largest supercomputer and it grows every year.
There still is team competition and if you go to the teams that compete in the upper slots and read there forums there power come from recruitment and retention of folders, they have active competitions between their members and prizes that go along with the competitions they all have folders that fold all classes of WU's, they do whatever they can to have fun folding, that is there thing not Stanford’s.
Just curious as to where you think the incentive is going to come from for people to upgrade or to build the 8 / 12 / 24 / 48 core systems if people got the same reward for duel core or GPU.
2 - SM H8QGi-F AMD 6xxx=112 cores @ 3.2 & 3.9Ghz
5 - SM X9QRI-f+ Intel 4650 = 320 cores @ 3.15Ghz
2 - I7 980X 4.4Ghz 2-GTX680
1 - 2700k 4.4Ghz GTX680
Total = 464 cores folding
Re: Bigadv points change
The reference to "folding elite" were neither my words or my opinion of the few big contributors that's why it had quotation marks, it's been used several times in this thread as a negative..... which would only be correct if those that can and do process these big projects consider themselves as such, and no, I don't think the big boys and girls contribute a lot of the work here, they perform a small but important role of returning big projects as fast as possible when required,Renix wrote:There's a saying, "what if you gave a parade and nobody came." The "elitist" label and attitude doesn't win friends or supporters in this endeavor and is rather disheartening to read to say the least. This isn't about GPU versus SMP. This is about the huge disparity within the SMP QRB for the very large units that is out of proportion to the other not so large SMP "as perceived by the general donor contributing population." What it is doing is causing folks to question whether it is worth their time and money to continue contributing for the other areas of the PG project. If the big systems contribute so much to the project, then why waste the resources of the other systems? Well one big reason for the small stuff is that if there wasn't the groundswell of public opinion created through all the "little people" to the project, then it's quite likely that the big organizations contributing the big money and support to the PG effort wouldn't be there to contribute it. So it is a real double edged sword for PG to manage the public relations aspect of this issue. Otherwise they wouldn't even be wasting their time on it. No body likes to have their face rubbed in it. But that is what is happening here, and people are taking notice. There isn't really any "team competition" with this project anymore as the whole point system is just totally out of whack. Yes, the big boys/girls contribute a lot of work here now, and it's all basically a game between them. Other folks come and go. Look at all the large teams and their membership rosters, the majority of their members don't contribute. On their forums a good number of the folks have moved on to other projects where there is more interest and competition among members and teams. This points issue is just accelerating that migration. But then, who cares............. the big machines will solve it all.
As far as teams and competition go, I don't think there's a team out there that would discourage a current member from putting lots of $$ into a system to process these larger Wu's, nor simply say "go away" to a prospective new member already running one or more, Looking at it another way, as an employee of a company you are paid "$ X" per hour for say 40hrs a week, if one day you stay and work overtime you may get "$ X" x 1.5, if you stay back even longer the same day that may go up to "$ X" x 2, that's normally done because of the time factor in getting a certain job done, that doesn't mean that the work done by others in their normal 40hr week is somehow of less value.
Re: Bigadv points change
I don't think anyone has seriously proposed that the same reward be given to dual cores or gpus. The reward has always been there to build faster rigs. It used to be that if you built a rig 2x as fast as the benchmark machine, you got 2x ppd and that was enough. Now with the QRB its 2.8x the ppd for 2x as fast.Grandpa_01 wrote: Just curious as to where you think the incentive is going to come from for people to upgrade or to build the 8 / 12 / 24 / 48 core systems if people got the same reward for duel core or GPU.
Re: Bigadv points change
I never said give the same reward. Re-read it. I said that the most recent very large SMP QRB was not in proportion to actual increase in system performance. Same basic argument that ChasR is stating about the very large SMP QRB. Try reading verbatim what is written. As for the vocal minority, that is on both sides of the fence in this issue.
Re: Bigadv points change
so that would be a vocal minority of a small minority ? (just kidding),Renix wrote:I never said give the same reward. Re-read it. I said that the most recent very large SMP QRB was not in proportion to actual increase in system performance. Same basic argument that ChasR is stating about the very large SMP QRB. Try reading verbatim what is written. As for the vocal minority, that is on both sides of the fence in this issue.
As the thread I originally started was merged with the larger thread about the points system, looks like we are of the same opinion,
the issue was that lots of donors considered the new big Wu's were earning too many points, but instead of just adjusting those before they left "beta",
the whole Bigadv project was revalued (read devalued) in the process, if indeed the whole Bigadv project's goals of big Wu's returned quickly has lessened in
importance to the science than when it was first released then indeed revalue them when needed.
-
- Posts: 1122
- Joined: Wed Mar 04, 2009 7:36 am
- Hardware configuration: 3 - Supermicro H8QGi-F AMD MC 6174=144 cores 2.5Ghz, 96GB G.Skill DDR3 1333Mhz Ubuntu 10.10
2 - Asus P6X58D-E i7 980X 4.4Ghz 6GB DDR3 2000 A-Data 64GB SSD Ubuntu 10.10
1 - Asus Rampage Gene III 17 970 4.3Ghz DDR3 2000 2-500GB Segate 7200.11 0-Raid Ubuntu 10.10
1 - Asus G73JH Laptop i7 740QM 1.86Ghz ATI 5870M
Re: Bigadv points change
Just curious but who said they had lessened in Value. To my knowledge that is not what was said.Jester wrote:so that would be a vocal minority of a small minority ? (just kidding),Renix wrote:I never said give the same reward. Re-read it. I said that the most recent very large SMP QRB was not in proportion to actual increase in system performance. Same basic argument that ChasR is stating about the very large SMP QRB. Try reading verbatim what is written. As for the vocal minority, that is on both sides of the fence in this issue.
As the thread I originally started was merged with the larger thread about the points system, looks like we are of the same opinion,
the issue was that lots of donors considered the new big Wu's were earning too many points, but instead of just adjusting those before they left "beta",
the whole Bigadv project was revalued (read devalued) in the process, if indeed the whole Bigadv project's goals of big Wu's returned quickly has lessened in
importance to the science than when it was first released then indeed revalue them when needed.
2 - SM H8QGi-F AMD 6xxx=112 cores @ 3.2 & 3.9Ghz
5 - SM X9QRI-f+ Intel 4650 = 320 cores @ 3.15Ghz
2 - I7 980X 4.4Ghz 2-GTX680
1 - 2700k 4.4Ghz GTX680
Total = 464 cores folding