Revised plans for BigAdv (BA) experiment.

[Nathan_P]

BA will end up being folded by big corporate donors only, the 30k cpu that VJ mentioned never showed up in points so either they do it for no points or they don't credit until they go public. That's where BA will end up, or at PDC if they can ever stump up the funds again.

I'm staying with SMP will still be good for 300k PPD and about 25-30 WU

 

[RoXQi3x]

I recall the QMD project - they were great point wu's, but you could only participate if you had an Intel cpu**! This project was announced very shortly after I had JUST bought two AMD based PC's for folding.

Thanks for this perspective, I really enjoy learning stuff like this!

 

[Audioaficionado]

This is why I build for myself and my own long term needs. I keep FAH in mind as I plan the build, but I don't build dedicated FAH machines. I've seen too many people invest $$$$+ in one trick ponies only to get depreciated with little warning.

I'll still keep folding as long as the power bill doesn't take out the food budget or doesn't seem to be contributing anything significant to the project to justify the extra expense.

 

[Jmtyra]

"The BA experiment will permanently end on January 31, 2015."
Bitcoin mining, here I come!

 

[Cobrabill93]

The BA change thread at the FF has been locked, and posts have been edited by the mods.

 

[orion456]

"The BA experiment will permanently end on January 31, 2015."
Bitcoin mining, here I come!

Wasn't the ceo of bitcoin just arrested? Though it does seem an interesting alternative for BA machines.

 

[Adak]

The BA change thread at the FF has been locked, and posts have been edited by the mods.

I'm not surprised. There was a lot of horse**** in that thread: inaccurate posts, posts bringing up issues with VJ and FAH, from W-A-Y back (years) ago, etc.

Vijay answered the questions that were asked, in a definitive way (namely that the BA project would stop). Of course, there will be those who don't like his decision, but that is to be expected, and an entirely different topic, which deserves it's own thread if it's to be discussed.

I haven't looked in on the thread recently, but when I was actively posting in it, the mods were giving the posters a lot of latitude in what they could post. There was a lot of highly emotional "venting" going on.

 

[Nathan_P]

I'm not surprised. There was a lot of horse**** in that thread: inaccurate posts, posts bringing up issues with VJ and FAH, from W-A-Y back (years) ago, etc.

Vijay answered the questions that were asked, in a definitive way (namely that the BA project would stop). Of course, there will be those who don't like his decision, but that is to be expected, and an entirely different topic, which deserves it's own thread if it's to be discussed.

I haven't looked in on the thread recently, but when I was actively posting in it, the mods were giving the posters a lot of latitude in what they could post. There was a lot of highly emotional "venting" going on.

Lets just say that the mods got bored with allowing us to have a discussion, and there are issues with FAH and how its run/administered, i'm not going to repeat them but here's the link, post #706 http://forums.evga.com/tm.aspx?m=2079048&mpage=24

 

[orion456]

Lets just say that the mods got bored with allowing us to have a discussion, and there are issues with FAH and how its run/administered, i'm not going to repeat them but here's the link, post #706 http://forums.evga.com/tm.aspx?m=2079048&mpage=24

Most interesting from that site for me was:

"so p5-133xl just posted:

I looked at the FLOP's numbers: http://fah-web.stanford.edu/cgi-bin/mai ... e=osstats2 and came to some startling conclusions.

1. 96% of the FLOPS are being done by GPU's and only 4% by all CPU's.
2. only .08% of FLOPS are done by Linux and that includes all bigadv machines.

My conclusions:

If the number of clients dropping is not a concern because the FLOPS are staying consistent then obviously people are dropping large quantities of CPU slots in favor of a few GPU slots in droves.

If the measure of scientific work done is the FLOP then CPU's should have little value and bigadv is equivalent to none. The bigadv program being dropped is totally insignificant to the F@H except at the level of PR.

PPD's do not accurately the measure of scientific work/value."


Which explains why BA is considered expendable.

 

[Cobrabill93]

Most interesting from that site for me was:

"so p5-133xl just posted:

I looked at the FLOP's numbers: http://fah-web.stanford.edu/cgi-bin/mai ... e=osstats2 and came to some startling conclusions.

1. 96% of the FLOPS are being done by GPU's and only 4% by all CPU's.
2. only .08% of FLOPS are done by Linux and that includes all bigadv machines.

My conclusions:

If the number of clients dropping is not a concern because the FLOPS are staying consistent then obviously people are dropping large quantities of CPU slots in favor of a few GPU slots in droves.

If the measure of scientific work done is the FLOP then CPU's should have little value and bigadv is equivalent to none. The bigadv program being dropped is totally insignificant to the F@H except at the level of PR.

PPD's do not accurately the measure of scientific work/value."


Which explains why BA is considered expendable.

If cpus are of so little value, the 260,000 SMP wus waiting on their servers to be folded are of little value too. Why the nudge off BA to SMP then?
Why would Bruce ask for help folding a backlog that may or may not even be? Why so many new SMP WUs being released? Why not just release new GPU Wus instead?
All this dose not add up.

 

[Adak]

If cpus are of so little value, the 260,000 SMP wus waiting on their servers to be folded are of little value too. Why the nudge off BA to SMP then?
Why would Bruce ask for help folding a backlog that may or may not even be? Why so many new SMP WUs being released? Why not just release new GPU Wus instead?
All this dose not add up.

It adds up just fine.

The SMP wu's ARE of just as much value as any other wu. Projects that started with SMP wu's, will be continued with SMP work units, to keep the data and wu's, consistent in format. I have never heard of a project that changed part way through the study, from SMP over to GPU format. GPU wu's require a much different code base.

I'm sure there was an excess of SMP work units - but FAH knows exactly how to stop that - quit making them worth so very little ppd, compared to other wu's.

Work units have NEVER been valued by the amount of science they produce. It's data, from a simulation. Nobody KNOWS in advance, which simulation will have the greatest value - or have no scientific value.

That's the nature of running an experiment, using the scientific method. A failure can be just as valuable as a great success -- or worth NOTHING. You don't know in advance, which result you will get, but even a "failure" is a step that leads to more knowledge.

GPU's have ALWAYS been powerhouse folders - far faster than the much more complex cpu's. HOWEVER, you need cpu's to feed the GPU, (although this is now being minimized), and you need computers with cpu's in them, in order to run the GPU, send in the results, get the next work unit, etc.

In the future, we'll be using GPU's, or dedicated cards that are similar to GPU's, to run these folding simulations. Today however, we still have a LOT of SMP work units to finish folding.

My opinion is that the BA program was too upset whenever VJ wanted to raise the threshold for the BA program. He patterned the BA program "like a super computer", but people are not super computer owners! To us, that's OUR money being used, and then VJ turned around and made it obsolete, all too quickly and/or too suddenly. Two increases in the BA threshold in just 60 days?? Talk about whiplash!!

 

[Cobrabill93]

Whiplash, that would be a pain in the neck. I think you are a couple feet to high.
The big iron BA server guys are hurting a couple feet lower!!!!!! LOL

 

[Adak]

Lets just say that the mods got bored with allowing us to have a discussion, and there are issues with FAH and how its run/administered, i'm not going to repeat them but here's the link, post #706 http://forums.evga.com/tm.aspx?m=2079048&mpage=24

There has ALWAYS been problems with how FAH is administered. Vijay had no prior experience working with a large volunteer donor project over the internet, before FAH. Also, his expertise is not in media relations. I cut him a lot of slack, for that reason.

I highly value the work that FAH is doing, and all of VJ's work to make it happen. His administrative work - there's definitely room for improvements there.

The big iron BA server guys are hurting a couple feet lower!!!!!! LOL

Joking aside, it actually took my breath away when he announced those two upgrades, so close together. I knew there would be changes down the road, but two jumps in the BA threshold, within 60 days?? Just stunning!

 

[orion456]

I knew there would be changes down the road, but two jumps in the BA threshold, within 60 days?? Just stunning!

Stunning yes - but also very stupid. The reaction was obvious. The dropping of the BA program probably will do more to damage FAH than just about any decision so far. It highlights the disrespect for the most dedicated donors. Those are people you don't throw away for no reason. On the other hand, maybe there is a reason and as usual we won't know about it until the next hammer falls. The lack of apparent planning in FAH is stunning indeed.

This is the 3rd time I've been stung -- and you know that saying....fool me once..blah blah blah.....

 

[Audioaficionado]

I'm sure there are plenty of decisions PG makes that they play close to their vest. There are other competing projects that also want grant money that they don't want to tip off to their methodology and results.

These decisions PO the $$$$+ dedicated donors for sure, but they are still learning what algorithms work as they go and make decisions based on their needs not ours necessarily. The competitive point system is a necessary evil to get the participation to advance the science.

I'm sorry to see the BA go away since I'll never get the chance to cash in by the time I have a system that will qualify. I'm going to have to eventually either upgrade or quit as I can no longer justify the high $$$ power bill for the meager 12kPPD I get.

 

[Jmtyra]

Wasn't the ceo of bitcoin just arrested? Though it does seem an interesting alternative for BA machines.

Haha, yes! I heard that in the news recently. It made me smirk.
TBH I'm not sure what I'll do...but I digress.