Folding@home

[pascal]

This is a very good idea: http://folding.stanford.edu/

I am running their client for Windows and it has no perceptible impact in SOHO activities. :)

Lets help it :mrgreen:

[zidane1strife]

Haven't seen the link, that's one of those protein folding mapping projects right?

If so then good idea.

[MrsSkywalker]

Thanks for the link, pascal! I hadn't heard of this project before.

[Crusher]

I use the other version: http://www.foldingathome.org/

[SvP]

3D enthusiasm and Folding are match made in heaven 8) Join Team Rage3D (http://rage3d.com/board/showthread.php?s=56a18a5b03d0a41958ff13ebfd2b2719&threadid=33617131) today! :lol:

Some promotional F@H videos (http://forum.folding-community.org/viewtopic.php?t=3770)

[pascal]

Haven't seen the link, that's one of those protein folding mapping projects right?

If so then good idea.Yes, the Stanford protein folding project :)

What about a Beyond3D Folding Team?

[DemoCoder]

This project will work alot better if we all had petaflop RSFQ desktops with holographic memory.

[Dave Baumann]

What about a Beyond3D Folding Team?

You wanna start one? :)

[pascal]

What about a Beyond3D Folding Team?

You wanna start one? :)Yeah. The FAQ is here: http://folding.stanford.edu/faq.html#stats.team

The creation form is here: http://folding.stanford.edu/cgi-bin/createteam
We need a team Name, Homepage and Logo (or just some URLs to start), then we can create the team.
My sugestion for the Team Name is B3D_Power :mrgreen:

Maybe in the future we could use the FP power of the GPUs to do the job. :)

[Tagrineth]

I use UD (www.uniteddevices.com).

[pascal]

A better name could be Beyond3D_Team

The list of the top 100 teams: http://folding.stanford.edu/cgi-bin/new/searchteamstats

[zidane1strife]

Hmmm, Do you think that processors like cell will make a difference(IOW significantly hasten/increase the speed) for this kind of thing in the near future, or will such processors be unsuitable for such purposes?

[bystander]

What about a Beyond3D Folding Team?

You wanna start one? :)

I would definitely be interested in joining a B3D Folding Team 8) so you have my vote. :)

[pascal]

Hmmm, Do you think that processors like cell will make a difference(IOW significantly hasten/increase the speed) for this kind of thing in the near future, or will such processors be unsuitable for such purposes?Probably it will work well, but the challenge will be program it (low level side). The high level algorithms work is already done. See below:Why not just use a supercomputer?

Modern supercomputers are essentially clusters of hundreds of processors linked by fast networking. The speed of these processors is comparable to (and often slower than) those found in PCs! Thus, if an algorithm (like ours) does not need the fast networking, it will run just as fast on a supercluster as a supercomputer. However, our application needs not the hundreds of processors found in modern supercomputers, but hundreds of thousands of processors. Hence, the calculations performed on Folding@Home would not be possible by any other means! Moreover, even if we were given exclusive access to all of the supercomputers in the world, we would still have fewer cycles than we do with the Folding@Home cluster! This is possible since PC processors are now very fast and there are hundreds of millions of PCs sitting idle in the world.

IIRC at any given time they have 35,000 CPUs available to do the work. :shock:

[Vince]

Hmmm, Do you think that processors like cell will make a difference(IOW significantly hasten/increase the speed) for this kind of thing in the near future, or will such processors be unsuitable for such purposes?

Absolutely, I couldn't help but chuckle when I saw this. Protein Folding is perhaps one of the most computationally expensive tasks that have yet been attempted and requires computational power that’s an order of magnitude greater (at least) than the ASCI series the DoE runs at massive costs. These people are wasting their time and resources that could better be allocated towards more established and less complicated tasks. (E.g. United Devices and their Anthrax/Cancer program)

an IBM researcher originally designed the Cellular concept for RT Ray-tracing, which never got the funding needed due to the small demand for such a device. Later, IBM resurrected the concept for it's Blue Gene Program which would leverage a pFLOP (in it's initial incarnation) of raw computing power based on the cellular architecture (e.g. along with it's eDRAM and speedy cache access) to explore Protein Folding - which is the true work-horse of the biological world.

Probably it will work well, but the challenge will be program it (low level side). The high level algorithms work is already done.

The problem is the immense computational resources necessary for folding protein. For example, I remember hearing in one lecture that a pFLOP devoted to just Protein Folding would take 2.5 years to simulate just 100microsec of folding mechanics/kinetics. The Stanford site states that a super computer would need "hundreds of thousands" of processors to keep-up with the PC users at home who are running a plethora of other background tasks and are using the same architecture with the same bottlenecks as any traditional Von Neumann architecture... I find this comment ignorant.

The programming of it will be relatively easy as there has been over 30 years of research on massive parallel computing to leverage.

The stuff IBM, SUN and the like will be putting out in a little more than a year will make all you've done during this time frame look like a waste. IBM alone is gearing up for a pFLOP+ scale system devoted to Folding in 2005.

IIRC at any given time they have 35,000 CPUs available to do the work. :shock:

And this is nothing compared to what will be done in the 2004-2006 timeframe. This is going to be like the Human Genome Project PartII where the 3 years of work put in at first can be computed by the end of the project in a month.

[Crusher]

Reminds me of http://www.fpx.de/fp/Fun/Googolplex/, where he proves that it is useless to begin running his program to print the number one googolplex anytime in the next 563 years, since the output would just be redunant.

However, just because protein folding will be done much faster in the future, it doesn't mean the results gathered today at a slower rate are not useful. The other folding project that I linked to has over 100 billion structures completed. That's a hell of a data set for any problem. They swap proteins every 10 billion, so they have 10 complete sets of data to work on right now. It's certainly giving researchers something to look at, even if it's slow. And that's with only a few thousand active users, not 90,000 active users like the Stanford project has.

[Vince]

The other folding project that I linked to has over 100 billion structures completed. That's a hell of a data set for any problem. They swap proteins every 10 billion, so they have 10 complete sets of data to work on right now.

For how long was the Folding Mechanics simulated? That what accuracy and were the true dynamics modeled? What type of Protein? What interactions are modeled and to what accuracy are Van der Waals and Lenard-Jones calculated? There are many things that raw nomenclature doesn't take into account, although big numbers seem to please everyone....

Saying they have a "Hundred Billion Structure" complete isn't saying much.

EDIT: Now that I actually read part of yout link Crusher, their not simulating anything are they? It appears that they're taking the Structure Prediction route and not simulating the actual folding dynamics or pathwath's. Hmm... no wonder their stating such big numbers if this is true, your just creating structures and attempting to find the proverbial 'needle' folding algorithm in the haystack of infinity (or it's like 20^40 or some insane amount IIRC). AFAIK, IBM dismissed this route for anything but refining.

[pascal]

Probably it will work well, but the challenge will be program it (low level side). The high level algorithms work is already done.

The problem is the immense computational resources necessary for folding protein. For example, I remember hearing in one lecture that a pFLOP devoted to just Protein Folding would take 2.5 years to simulate just 100microsec of folding mechanics/kinetics. The Stanford site states that a super computer would need "hundreds of thousands" of processors to keep-up with the PC users at home who are running a plethora of other background tasks and are using the same architecture with the same bottlenecks as any traditional Von Neumann architecture... I find this comment ignorant.

The programming of it will be relatively easy as there has been over 30 years of research on massive parallel computing to leverage.

IIRC they are working with 10microsec. Also the agregated pFLOP of 35,000 computers maybe more than 1 pFLOP for very low cost (operation and programming).

Also the number of computers depends on the number of people that want to help.


IIRC at any given time they have 35,000 CPUs available to do the work. :shock:

And this is nothing compared to what will be done in the 2004-2006 timeframe. This is going to be like the Human Genome Project PartII where the 3 years of work put in at first can be computed by the end of the project in a month.
Maybe but lets help Folding@home for now :wink:

[pascal]

The Beyond3D_Team was created. See it here: http://www.beyond3d.com/forum/viewtopic.php?t=5668

[Vince]

IIRC they are working with 10microsec. Also the agregated pFLOP of 35,000 computers maybe more than 1 pFLOP for very low cost (operation and programming).

That was my point Pascal, I think your doing a great and noble thing putting a team together and I'll support you; Yet, I don't see the point. Even with 35,000 users who are doing other things concurrently - we'll pretend there is 100% utilization (which is ridiculous when you think of the ineffeciencies piled on: Between the inherient Von Neumann architecture's problems and the additional user created ineffeciencies concerning the OS and aggregate preformance deminishing programs) that output an average 8GFlop/s. Thats ~280TFlop at 100% effeciency if I did the math in my head right. Thats a quarter of ONE Blue Gene system in theory, perhaps a 10th (~28TFlops) atmost in reality. Add to this that within 2 years you'll potentially be buying PCs and Consoles with TFlop range MPUs based on the IBM/SUN (or Intel to a lesser extent) type architectures that just scream massive concurrency and onboard DRAM; and buy them them for $300-$1000.

Add to that the simulation time of only 10msec and I qustion it's utility.

Also the number of computers depends on the number of people that want to help.

This be true, and that numbers goin' up by atleast two thanks to you :)

[pascal]

IIRC they are working with 10microsec. Also the agregated pFLOP of 35,000 computers maybe more than 1 pFLOP for very low cost (operation and programming).

That was my point Pascal, I think your doing a great and noble thing putting a team together and I'll support you; Yet, I don't see the point. Even with 35,000 users who are doing other things concurrently - we'll pretend there is 100% utilization (which is ridiculous when you think of the ineffeciencies piled on: Between the inherient Von Neumann architecture's problems and the additional user created ineffeciencies concerning the OS and aggregate preformance deminishing programs) that output an average 8GFlop/s. Thats ~280TFlop at 100% effeciency if I did the math in my head right. Thats a quarter of ONE Blue Gene system in theory, perhaps a 10th (~28TFlops) atmost in reality. Add to this that within 2 years you'll potentially be buying PCs and Consoles with TFlop range MPUs based on the IBM/SUN (or Intel to a lesser extent) type architectures that just scream massive concurrency and onboard DRAM; and buy them them for $300-$1000.

And what about hundreds of thousands of nodes and each one with the MPUs you described? :shock:


Also the number of computers depends on the number of people that want to help.

This be true, and that numbers goin' up by atleast two thanks to you :)Thanks to all of us :)

[Crusher]

More in depth, detailed information on the background and theory:

http://bioinfo.mshri.on.ca/trades/index.html

[pascal]

zidane1strife
MrsSkywalker
Crusher
DemoCoder
SvP
Tagrineth
bystander
Vince

Lets move. The Beyond3D_Team number 32377 is waiting you :wink:
http://www.beyond3d.com/forum/viewtopic.php?t=5668