Repetition of long standing and continuously failing policy?
Yes.
Macholic posts:
Old horse. New year is coming.
Crap Points and GPU WUs
So, I quit Folding@Home
Unbalanced Scoring
Points are (or should be) social engineering
Uniprocessor client or SMP?
[Debate] Revamped points system based on early return of WUs
benchmarking F@H
F@H and World Community Grid
Comparing performance between folding clients
Curious, why are points so important to some?
Open letter to PG [with response]
What is going on with BIGADV?
Fair distribution for PPD BigAdv / GPU
Is PPD adjusted for inflation as time passes? [No]
Lopsided points when folding -smp
Pande Group announcement about Big Advanced Projects
Suggested Change to the PPD system
PG's use of points to influence how the science is done
Bigadv Reform
Blog post "Unified GPU/SMP benchmarking scheme..."
That is in this forum section alone. Out with the old, in with the new. Happy New Year!
The implication being that complainers gonna complain...or that the problem has been discussed ad nauseam, so why bother making any attempt to address the problem now?
Kasson posts
We do pay attention to feedback. We may not respond, but we do listen. We appreciate that this is a complicated problem. And we appreciate the participation of all of our FAH donors.
Well, here might be the problem.
Perhaps several dozen threads on the subject indicates that a different solution method might be required.
Perhaps instead of passive reactionary (or non-reactionary) listening...
Perhaps communication is a two way street?
Perhaps pro-active, anticipatory sharing of information and discussion of pending issues, backlogs, or diminishing requirements for certain WU would be helpful.
Perhaps more participation in the discussion and less passive listening is the key to solving this "complex" problem.
Just an idea.
Transparency and Accountability, the necessary foundation of any great endeavor!
Question do any of you think that in reality PG is going to get there share of donor support if there is not some kind of consistency and balance along with donor appreciation and support. There are more problems than just points, much of which generate from this very site. Many of you read the other forums and know exactly what people think of what happens here. And it is not good.
When a large % of you contributors think there is a problem then there is most likely a problem, the current problem appears to be the difference in point values between projects bigadv = xxxxxx ppd which drives members to buy the equipment and pay the power bills needed to reach xxxxxx amount of ppd. $$$$ = points.
GPU core 17 = xxxxx ppd which again drives members to spend the $$$$ to buy the equipment and pay the power bills needed to reach xxxx again amount of $$$$ spent = ppd
SMP = xxx ppd = Hmmm I can not justify it I can buy a GPU and make xxxxx or a MP system and make xxxxxx so I am goig to spend where I can do the best = $$$$ per ppd.
So yes no matter how you slice it $$$$$$ = science done it is already there and always has been.
I myself have been a long time donor to Stanford and my contribution from the early days has been devalued as I expect my current contribution to be devalued, ever hear of inflation it is an every day part of life. And yes I remember some of the names from the early years but how many of the current folders have even seen there ames and why should they. Living in the past is not going to make a very bright future.
If PG decide to stay in the past that is where they stay O-well that is up to Vj and crew but the reality is that if you do not grow and progress with time you will get left behind, but as I said earlier there is allot more wrong here than just ppd, and that needs to be dealt with also.
I saw Kasson's post that the requirement is changing. I won't be spending money to keep up with the Jones', as the equipment no longer makes BA, it will be retired/donated/sold. There is still hardware out there that can make the deadline, but, it has been or will be obsoleted.
V7 GPU client has proven to be too much of a chore to keep running in comparison to v6 gpu/ba clients. I don't fold on my gpus. Any complaints on the V7 client turn into a few(the two) of the trolls around here coming out and being worthless as usual.
I guess it is a matter of time before I no longer fold.
There have been many posts here, so I'll try to address the points raised by multiple posts.
1) Concept of equal work. There's a difference in how some donors and PG view equal work. In many posts, donors are thinking of equal work in terms of Watts. I understand why, but that's not equal work from the perspective of how much science is accomlished. Equal work for us means the same calculation performed at a given time (i.e. we don't expect points awarded now to be identical to points 12 months ago). Our general philosophy is that we reward points based on the science done (not the Watts expended) and give the donors the choice of how they would like to perform that work (i.e. SMP vs GPU). Our thinking here is that donors know best how they want to run FAH and we should leave it to them to optimize whatever they want to, if they can do the same science.
2) Availability of Core17 WUs. We have several Core17 projects that we are ramping up. Core17 is a key part of how we will be getting calculations done in the future. If there are issues with getting Core17 WUs, please start a thread for that issue specifically so we can get people on that to fix it. If there are issues with the GPU client, please also start a thread for that as well. I think proteneer has been working hard to be responsive to donors concerns and questions about Core17, so please just let us know what's on your mind there.
3) The nature of bigadv. Bigadv has always been intended as an experimental, ultra-extreme requirement subset of FAH calculations and FAH machines. We've tried to set the expectations that the requirements not just can change but will change periodically as computer hardware progresses. It's clear that that expectation hasn't been communicated well and without that, I can see how donors would be upset, and that makes me worried that the bigadv experiment has been handled poorly on our part. Moreover, from these posts, I am also worried that it is just intrinsically difficult for us to have such an experiment running and still maintain fairness to all donors (those running bigadv and those not). Considering that the requirements for bigadv will be increasing in time, maybe what would be best is for us to try to lay out further in advance the schedule for changes? It's hard for us to predict too far in the future, since that essentially means predicting changes in computer hardware, but it sounds like at the very least we should be giving more heads up to donors for changes expected. In that spirit, beyond the ramp up Dr. Kasson discussed, I expect that there will be another change in requirements closed to the end of 2014, to keep up with changes in hardware.
4) Donor communications. Finally, we had an external group come in and ask donors what they thought could be best improved in FAH and donor communications came up strongly. I've been thinking about what we can do to help improve that in a sustainable way. We've had many attempts but keeping the communication going is probably only possible with us having someone who's primary job is donor relations and communications -- the science and development team are as it is overloaded with the tasks of keeping FAH running. So, I have been looking into hiring (into a part time role) someone who's sole job is donor relations and communications. Hopefully that will help communication in both directions. If not, we'll hopefully learn from that and continue to see what we can do to get better.
Sorry for the long post and for all the grief here for the bigadv issue. We're continuing to talk internally to see how we can improve things based on your comments.
Happy holidays and fold on!
Prof. Vijay Pande, PhD
Departments of Chemistry, Structural Biology, and Computer Science
Chair, Biophysics
Director, Folding@home Distributed Computing Project
Stanford University
BA was not drawn up with the average F@H user in mind. The hardware to run BA properly has always been a 2p machine or greater which 99.9% of users do not have at home. PG expressed a desire for this style of hardware by releasing these project and the serious folders who were able to afford to went and invested a lot of cash in the systems that currently run BA units. To have these users suddenly have their hardware obsoleted is unfair to say the least. Every machine out there running BA units returns its unit within 2.4 days or less unless there is a problem somewhere - this is not like the old days of spoofed i5's, X6's and i7's that barely made the grade and more often than not didn't. These days its (expensive) server grade hardware that will run for years on BA units with no hassle - Hell I bet some of the top end machines return BA units faster than a lot of SMP units get returned
These users are the same ones that are usually the most active and most helpful towards the rest of the F@H community. It is going to be a serious loss to the project as a whole when/if these users start shutting down.
Analogy, I invest X to get Y return - if Y is suddenly 40%/50%/60% less what would you do - especially when it costs real money??
BA only gets done because we - THE DONORS - are willing to spend the cash required to buy the machines needed to do the job - I do not need 5 2p machines for day to day use
1:- Equal pay for equal work:- My x5670 machine is capable of the same work in May that it is capable of now, nothing has changed on the hardware front, it gets 115-170k PPD depending on the project. From May an arbitrary rule comes into force that says my machine is no longer capable of equal work. How does that work? A quad core now is the same as a quad core from 2007 - it still runs smp but may be quicker. The older machines don't get relegated to having to run 4 uni proc clients do they - this is what is happening with BA.
2:- No comment, I don't run gpu.
3:- As in my earlier post. BA requirements are not "@home". not many people need a multi cpu machine for day to day use. PG set this up knowing that donors with cash to burn would invest the necessary money and time to get machines that will do the job. To continually have the rug pulled out from under us is unfair to say the least. Our cash saves PG from having to rent supercomputer time or have the projects broken up into smp friendly lumps. When it gets to the point that consumer class hardware can run BA units then look at it. BA folders are a small bunch of very dedicated donors - turn us away and BA won't get done - period
4:- I'll reserve comment until the donor relations person gets appointed.
additional, a 3rd set of changes in 2014 would probably rule out all but 48/64 core G34 and lga 2011 hardware + the odd lga 1567 machine, even lga 1356 won't be good enough
Due to my own financial circumstances I haven't folded BA since the threshold was increased from 8 cores, so I'm not really in the 'affected' group. I've also understood that BA was a moving target (and that the higher point premium is justified by that). I do like the tentative extended forecast and would love to know what factors Pande Labs will use to know when it's time to make a subsequent change to the threshold.
One thing I would have really liked to see in the blog post was some of the metrics/data behind the decision. Perhaps it is only triggering my sense of curiosity, but I think that is quite beneficial to helping donors feel part of the project.
Little things like:
How much has the BA pool changed since the last revision?
Have we seen a major increase in the number of 24 and 32+ clients?
What was the threshold that triggered the change in threshold and by how much is BA (and perhaps non-BA SMP) going to be more effective as a result following it? (Presumably higher average return more than offsets the reduction in parallel clients)
How well does core count translate to performance?
Threshold change is a 'shock' to the system - how has each shock changed the effectiveness of BA in the past?
VijayPande wrote:Considering that the requirements for bigadv will be increasing in time, maybe what would be best is for us to try to lay out further in advance the schedule for changes?
Advance notice is much appreciated. Though, given significant difference between same-core-count hardware, it's not of much practical use.
What bigadv donors actually need (in context of bumping bigadv requirements) are changes to preferred deadline times (if any) effective, respectively, February 17 and April 17.
Only that will allow proper planning on bigadv community part.
If you could facilitate obtaining that information, it would be awesome.
Now that we are actually hearing from people at PG (Vijay and others), I feel that I can offer my thoughts in an environment where it may be heard.
Dr. Pande, I very much appreciated that you stepped-in and are speaking to us about this subject. I feel that many Folders (that are willing to invest in these Bigadv machines) would be willing to help you with the SMP WU needs if someone would ask us first. Sure we like to see our Bigadv rigs working Bigadv WUs, but I think many of us would be just as willing to help you guys out. As Nathan shared, these Bigadv machines are expensive to build. Unless someone has invested thousands into one of these Bigadv machines, they really can't understand the pain that sets-in when you are surprised to find that it will no longer be relevant to Bigadv in a few months. I have several thousand dollars invested in my (2) Bigadv rigs (a 2P and a 4P).
There are limits (as I'm sure you can appreciate) to how far people can (or are willing) to invest in doing Bigadv work. It would be especially helpful if your group could give us a better roadmap of what you expect (now and in the future). Regular updates on that roadmap would also be useful to prevent shock and the resulting donor disappointment that comes when a donor realizes in 2 months their machine will be obsolete to Bigadv work.
I work for the largest computer company in the world (Mainframe division) and know that our customers expect us to provide roadmaps on our Mainframe systems. High-end investors in Folding equipment are not much different. They want to know how much time their investment in a given Bigadv platform will last and to know what may impact that investment going forward. I'd like to see your group provide some very specific information as to what will be required. If at the end of 2014, a Bigadv Folder would need to have a $5-10,000 system to Fold Bigadv, then people need to know that so they can make good decisions. It is one thing to say a GPU has to be replaced because it can no longer Fold a given WU. It it quite another thing when it is a Server class machine, costing thousands of dollars that is all of a sudden obsolete to Bigadv. 2 months warning is minimal in my view and needs to be much longer considring this kind of investment.
Lastly, I think it would also be good to "open the kimono" and tell Bigadv Folders how your group sees our future participation related to Bigadv. Are you needing much higher-end servers to tackle Bigadvs at x-time with x-power? If we know these things, honestly, openly and from your group (not other Folders here), we could be better informed about whether to even get involved in Bigadv going forward.
Hardware configuration: Intel i7-4770K @ 4.5 GHz, 16 GB DDR3-2133 Corsair Vengence (black/red), EVGA GTX 760 @ 1200 MHz, on an Asus Maximus VI Hero MB (black/red), in a blacked out Antec P280 Tower, with a Xigmatek Night Hawk (black) HSF, Seasonic 760w Platinum (black case, sleeves, wires), 4 SilenX 120mm Case fans with silicon fan gaskets and silicon mounts (all black), a 512GB Samsung SSD (black), and a 2TB Black Western Digital HD (silver/black).
Dr. Pande, please confirm, that like last time, that the core count threshold is only going up on new projects released after those due dates, and that current projects will continue as is until they are completed (weeks, maybe months later).
Please confirm whether the deadlines will be adjusted or not adjusted.
Please consider asking all the BA folders to fold SMP work units for the next month, and turn off the BA assignments while that's done. Then maybe you wouldn't have to change the BA requirements at all, or that could delay the need for a change for a good while.
7im wrote:Dr. Pande, please confirm, that like last time, that the core count threshold is only going up on new projects released after those due dates, and that current projects will continue as is until they are completed (weeks, maybe months later).
Please confirm whether the deadlines will be adjusted or not adjusted.
Please consider asking all the BA folders to fold SMP work units for the next month, and turn off the BA assignments while that's done. Then maybe you wouldn't have to change the BA requirements at all, or that could delay the need for a change for a good while.
I think the folders will respond to a call to arms so to speak to get rid of any backlog of SMP WUs.
All you have to do is ask. The folding community is a generous group and will help.
This is the firs day I haven't folded in seven years. That's how I feel about it. I spent thousands this year alone.
You did not ask me to, but still that is real money I could of ....................................... with.
