Change in BA requirements

[mdk777]

Well, I have nothing against the efforts any have made...either as official or unofficial spokespersons...or just attempting to influence public opinion.

My view is that "self-regulation"... allowing this forum to run with a hands off approach...just has not proven sufficient.

It is no reflection on any individuals personal effort, but rather a reflection on the flawed structure of the program.

While I see a consensus among many here, I see that consensus as fatally flawed.

a push to reduce the backlog of smp WU does nothing to improve the understanding of how BA is being administered now, or will be administered in the future.

Actually making assumptions and educated guesses is what brought the current discussion to start.

It was just as logical of guess that the BA program would continue in its current form, as it was to assume there was some pressing imbalance that needed to be redressed.

I have said this dozens of times over the years:
We can discuss and theorize, and calculate and suppose for infinity.
What is really needed to permanently resolve these issues is leadership and clarity from PG.
All our opinions might well be barking up the wrong tree. We make baseline assumptions and extrapolate.

Bruce assumes that smp projects are falling behind.
I note that the projects that actually require and benefit from BA might be in low demand at the moment.
We all note that open mm 5.2 has not been implemented yet and that GPU work units might be awaiting a revision in core17

My point:
Lets eliminate the middleman!
lets just have clear and open communication, and we can all stop our hobby of trying to guess what is the real rational for decisions.
We can then all apply for jobs in the CIA as analysts projecting the next move of the regime on N. Korea.

[Bill1024]

QRB is based on how fast a WU is returned. A 24, 32, 48 or a 64 core should be sending them back real fast.
More so than a Q6600 would. Just figure a way to make it so a 48 4P will be close to PPD doing either SMP or bigadv.
Say 2 x base value for 24 hours, 3 x for 18 hours, 3x for 12 hours, 4x for 8............8x for 2 hours, bonus 20 x for under 30 minutes.

Grandpa wrote. (Sorry I did not edit it exactly right)
Not even close you get about 1/2 the PPD on the best smp vs bigadv and around 1/3 the PPD on the worst. that is a difference of 350k to 500k PPD on my rigs

You are right, I now see what my 48 core can do BA vs SMP. It is a huge point drop for the same powerful computer and electric power usage. I also remember from last time 2 years ago.
The K factor is set at .75- 3.xx and that is fixed right? What I am trying to say is, "make that k factor (the multiplier) a sliding scale so as return times go down, the multiplier (K factor) goes up".
So a 48 core may return a SMP in a couple hours and have a 20 k factor (multiplier). To the point a 4P can make near as much doing either WU. You may not want them exactly even.
But a lot closer is in order. IMHO And many here agree.
A slower 1P 4 core may get a 3x, a 6 core AMD 1045T may only get a 5x, a 1099T a 6x. A dual quad Intel 8/16 a 10 K factor. Faster return better points.
I am in no way saying to change anything about the points on BA.
Those numbers would of course be tested and adjusted.
For sure the SMP points are low and this is why donors do not want to fold them. I do not know if it is possible to make a sliding scale.
Does anyone think it can make this situation better?

[Grandpa_01]

Upcoming changes to bigadv threshold
by kasson » Tue Dec 17, 2013 8:43 am

We have a policy of periodically re-evaluating the bigadv program, including the threshold required to run bigadv projects.
It is the intent of bigadv to match large and resource-intensive work units with some of the most powerful machines used by FAH donors. This "most powerful" line naturally advances with computing power. To date, bigadv has been a CPU-based program, and with increasing numbers of CPU cores and power of those cores, we have decided to lay out a roadmap of bigadv threshold changes for the next several months.

Feb 17 (two months from today): bigadv threshold will become 24 cores
Apr 17 (four months from today): bigadv threshold will become 32 cores

We want to give advance notice of these changes, and we recognize that change is not always welcome or comfortable. We should also emphasize that the science performed by donor machines is valuable in all parts of the FAH project, and part of the change in bigadv threshold is because we would like to encourage moderately powerful machines to help boost the capabilities of non-bigadv SMP projects where we do a lot of this science.

We also recognize that core count is not the most robust metric of machine capability, but given our current infrastructure it is the most straightforward surrogate to evaluate.

It appears to me many are reading more into this than is intended, at least that is the way it seems to me and many have gone way out of line in providing answers to something that was never asked for by Stanford. What is happening is nothing more than human emotional reaction generated from an incomplete statement. I do not see anywhere that it says there is a huge backlog or that they want bigadv capable rigs to switch to smp at this time.

We have a policy of periodically re-evaluating the bigadv program, including the threshold required to run bigadv projects.
It is the intent of bigadv to match large and resource-intensive work units with some of the most powerful machines used by FAH donors. This "most powerful" line naturally advances with computing power. To date, bigadv has been a CPU-based program, and with increasing numbers of CPU cores and power of those cores, we have decided to lay out a roadmap of bigadv threshold changes for the next several months.

Feb 17 (two months from today): bigadv threshold will become 24 cores
Apr 17 (four months from today): bigadv threshold will become 32 cores

I see nothing here that says they are doing anything other than attempting to lay out a road map for future bigadv qualifications. Could have this road mapping have been done better ? (YES) hopefully they are working on that. But they are doing what they said they were going to do several months back and attempting to give a heads up. And let’s face it Bigadv was and is designed for the most powerful machines and unfortunately most of the 2P rigs do not fall into that class.

We want to give advance notice of these changes, and we recognize that change is not always welcome or comfortable. We should also emphasize that the science performed by donor machines is valuable in all parts of the FAH project, and part of the change in bigadv threshold is because we would like to encourage moderately powerful machines to help boost the capabilities of non-bigadv SMP projects where we do a lot of this science.

I do not see where it says they need us to switch over to smp from bigadv or anything else for that matter, This is just shear panic caused by people not reading what was written and listening to others who did not read what was written it is the proverbial (snowball rolling down hill) who on earth from PG asked us to switch our machines to smp if Stanford wanted more machines running smp they will take the appropriate steps to either entice us or force us. Hopefully not the second I doubt that would accomplish the desired effect.

[Viper97]

I'm still waiting to hear from the big guy himself. Meanwhile this is a good time for me to do some preventive maintenance on my rigs. I'm setting one boot stick to run strictly WCG and another for folding.

My main point is I suppose that other than VJ's post on the website about new processes and improvements. I don't see a whole lot of 'authorized' spokes persons representing Stanford.

[Bill1024]

and part of the change in bigadv threshold is because we would like to encourage moderately powerful machines to help boost the capabilities of non-bigadv SMP projects where we do a lot of this science.

The problem that has been talked about, and exposed here is, "the PPD for SMP STINKS".
A 4P 24 core server class computers PPD AS Grandpa said is 1/3 - 1/2 on SMP, what it can get doing the BA.

I am very good at reading between the lines. My wife holds her hand up to me all the time and tells me to" read between the lines."
I know what she is saying to me. (She is saying she loves me)

I read that statement to say we are going to nudge you off bigadv to fold SMP.
There must be a reason and I bet it is because the smp are not getting done. People with 6 or 8 core are not folding smp with their GPUs
I always hear them say "It's not worth the electricity"
In a healthy economy when there is full employment and every one has a job; the employer has to pay more to get people to work digging ditches or washing dishes, or working at McD's.
If there is a type of work that is not getting done then you have to pay more, simple economics.

[mdk777]

yeah, maybe I'm dense.
just can't conceive of a more irrational and dis-motivational way of going about achieving that goal.

[Grandpa_01]

and part of the change in bigadv threshold is because we would like to encourage moderately powerful machines to help boost the capabilities of non-bigadv SMP projects where we do a lot of this science.

The problem that has been talked about, and exposed here is, "the PPD for SMP STINKS".
A 4P 24 core server class computers PPD AS Grandpa said is 1/3 - 1/2 on SMP, what it can get doing the BA.

I am very good at reading between the lines. My wife holds her hand up to me all the time and tells me to" read between the lines."
I know what she is saying to me. (She is saying she loves me)

I read that statement to say we are going to nudge you off bigadv to fold SMP.
There must be a reason and I bet it is because the smp are not getting done. People with 6 or 8 core are not folding smp with their GPUs
I always hear them say "It's not worth the electricity"
In a healthy economy when there is full employment and every one has a job; the employer has to pay more to get people to work digging ditches or washing dishes, or working at McD's.
If there is a type of work that is not getting done then you have to pay more, simple economics.

I like reading between the lines also but what that says is a few months down the road. They are not asking for people to go off bigadv and onto smp, if there was a need they would have said we are changing the requirements today, that is simply an attempt at a roadmap.

Reading between the lines can sometimes create more harm than good.

[7im]

Okay, I'll play dumb. Where does that roadmap lead to a few months from now?

[Bill1024]

and part of the change in bigadv threshold is because we would like to encourage moderately powerful machines to help boost the capabilities of non-bigadv SMP projects where we do a lot of this science.

The problem that has been talked about, and exposed here is, "the PPD for SMP STINKS".
A 4P 24 core server class computers PPD AS Grandpa said is 1/3 - 1/2 on SMP, what it can get doing the BA.

I am very good at reading between the lines. My wife holds her hand up to me all the time and tells me to" read between the lines."
I know what she is saying to me. (She is saying she loves me)

I read that statement to say we are going to nudge you off bigadv to fold SMP.
There must be a reason and I bet it is because the smp are not getting done. People with 6 or 8 core are not folding smp with their GPUs
I always hear them say "It's not worth the electricity"
In a healthy economy when there is full employment and every one has a job; the employer has to pay more to get people to work digging ditches or washing dishes, or working at McD's.
If there is a type of work that is not getting done then you have to pay more, simple economics.

I like reading between the lines also but what that says is a few months down the road. They are not asking for people to go off bigadv and onto smp, if there was a need they would have said we are changing the requirements today, that is simply an attempt at a roadmap.

Reading between the lines can sometimes create more harm than good.

They are not asking moderately powerful machines to go off bigadv. They are pushing them off bigadv, and encouraging moderately powerful machines to boost the capabilities of SMP.

OK we are parsing words here.
Can someone from PG define what boost the capabilities means exactly Please?
And make clear, once and for all, is there a need for more CPUs, to be directed at getting SMP WUs done, because of some "perceived" backlog that may or may not exist.

[Bill1024]

and part of the change in bigadv threshold is because we would like to encourage moderately powerful machines to help boost the capabilities of non-bigadv SMP projects where we do a lot of this science.

The problem that has been talked about, and exposed here is, "the PPD for SMP STINKS".
A 4P 24 core server class computers PPD AS Grandpa said is 1/3 - 1/2 on SMP, what it can get doing the BA.

I am very good at reading between the lines. My wife holds her hand up to me all the time and tells me to" read between the lines."
I know what she is saying to me. (She is saying she loves me)

I read that statement to say we are going to nudge you off bigadv to fold SMP.
There must be a reason and I bet it is because the smp are not getting done. People with 6 or 8 core are not folding smp with their GPUs
I always hear them say "It's not worth the electricity"
In a healthy economy when there is full employment and every one has a job; the employer has to pay more to get people to work digging ditches or washing dishes, or working at McD's.
If there is a type of work that is not getting done then you have to pay more, simple economics.

I like reading between the lines also but what that says is a few months down the road. They are not asking for people to go off bigadv and onto smp, if there was a need they would have said we are changing the requirements today, that is simply an attempt at a roadmap.

Reading between the lines can sometimes create more harm than good.

They are not asking moderately powerful machines to go off bigadv. They are pushing them off bigadv, and encouraging moderately powerful machines to boost the capabilities of SMP.

OK we are parsing words here.
Can someone from PG define what boost the capabilities means exactly Please?
And make clear, once and for all, is there a need for more CPUs, to be directed at getting SMP WUs done, because of some "perceived" backlog that may or may not exist.

Where exactly Bruce stands on the ladder in the chain of command I do not know.
But he did ask for help and to take aim at the regular SMP wu,s
I did not see any one else from PG say that is NOT a good idea and please do not do that at this time.

[Grandpa_01]

Okay, I'll play dumb. Where does that roadmap lead to a few months from now?

Well you would have to play dumb to follow the current road mapping scheme : lol:

But let’s give it a shot

In a few months from now it leads down the same path that it lead to a few months back a bunch of unhappy customers not wanting to or willing to fold smp until the value is raised and it becomes worthwhile. It also leads to an reevaluation of the bigadv qualifications (Enter Dates Here), (HINT) hopefully with deadline numbers rather than core count.
The problem of not enough smp folders is not going away until the smp value is raised or they port everything over to GPU which currently seems to be a bit of a problem. I do not advocate lowering anything as all that is going to do is create more strife.

There is another possibility but we will wait till we see if some of the announced changes come or not before we go down that path.
So for now we will just play dumb a follow the current scheme

Where exactly Bruce stands on the ladder in the chain of command I do not know.
But he did ask for help and to take aim at the regular SMP wu,s
I did not see any one else from PG say that is NOT a good idea and please do not do that at this time.

Bruce is the same as you me or anybody else here at the FF when it comes to PG he has stated it many times in the past, as far as the FF goes he is the big dog on the block. He is also a long time contributor and has been loyal to PG and the FF for many years. He can make educated guesses and read between the lines also. Sometimes he is right and sometimes he is wrong just the same as the rest of us.

[ChristianVirtual]

A roadmap would lead to a better understanding on what will come next ...

1) how many new projects are in the pipeline for what core (specially with respect to different GPU category)
2) what kind of new core/clients are in the pipeline; what key requirements (Xeon Phi, game consoles, cloud, mobile ...)
3) what kind of collaboration with hardware vendors (if any) take place ?
4) less important, but also of interests: development/release cycles for FAHclient

Just anything which supports donors decisions on next investments and where to focus on.

Example: If a potential GPU-BA would require 4 GB VRAM then I better buy now a GTX Titan instead of a GTX 780Ti. Or if SMP-BA gets an increase to core count/decrease in deadlines that influence what kind of system to build next or invest in different categories instead.

Like in regular drug development when a new substance not make it out of clinical trial; bad luck. Same if there is a plan to use JIT compiler from NV but they not finish it or Apple don't get the OpenCL sorted out. Nothing PG can influence too much. But they could adjust the roadmap and share with us the information they have (and not potentially protected by any NDA).
A good job for a communication/bridge person between PG and donors.

[Nathan_P]

Well here is one for you to put everything in perspective, I was just about to upgrade a rig at a cost of about £700 to be able to continue to do BA WU quicker and at less cost to me (lower electric consumption). Given the announcement that the requirements are going to change 3 times in 2014 I have decided to repurpose that £700 to other things that I need and when the changes come in April the current rig will be shut off.

Nice incentive, given that not many of your donors will spend that much to do BA work. Oh and as stated BA work is not @home using spare cpu cycles. One does not need a 2p/4p machine to surf the web and play games. Argue all you want, BA IS NOT @HOME, ANYONE WHO THINKS OTHERWISE IS LIVING IN A DREAMWORLD, PEOPLE INVEST LARGE SUMS TO BENEFIT SCIENCE. WITHOUT US BA IS DEAD.

Oh and more people run BA on 2p than 4p machines........

[PantherX]

...4) less important, but also of interests: development/release cycles for FAHclient...

Maybe this will help you somewhat:
1) Tickets sorted by last change -> https://fah.stanford.edu/projects/FAHClient/report/25
2) Active tickets by milestones -> https://fah.stanford.edu/projects/FAHClient/report/3
3) Roadmap -> https://fah.stanford.edu/projects/FAHClient/roadmap

I personally stick with the first one most of the time as it covers all my needs.

[Nathan_P]

OK, a simple question

The current donor pool for BA is small using the following hardware:- 4p is 4p G34, 4p LGA 2011, 4p socket F and the odd 4p LGA 1567. 2p is 2p G34, 2p LGA 2011, 2p 1366, 2p 1356 and the odd 2p C32.

The new core increase will remove 4p socket F, 2p LGA 1356, 2p LGA 1366, 2p C32, some 2p LGA 2011 and some 2p G34. How are less machines going to do more science?