SCE Joins Stanford's Folding@home Program (B3D ID=32377)
- Thread starter one
- Start date
IIRC, all discoveries are for the public domain with F@H.
Yes, as I understand it Folding is one of the very few undertakings that are undertaken for the greater advancement and benefit of humanity, furthering our knowledge and sharing that. So much ordinary science is led by agendas that it's worth is terribly depleted.
Jawed
Legend
http://gizmodo.com/gadgets/home-ent...more-powerful-at-folding-than-a-pc-244528.php
Whoops, shoulda also posted the video:
http://gizmodo.com/gadgets/home-ent...-home-app-churning-away-at-protein-244534.php
Jawed
Whoops, shoulda also posted the video:
http://gizmodo.com/gadgets/home-ent...-home-app-churning-away-at-protein-244534.php
Jawed
I don´t hink it will ever happen, decoding MPEG4 and better at 1080p is a lot of work.
Rainbow Man
Veteran
Cell should have several times the basic grunt neccessary especially considering some steps of the process can be dumped over onto the GPU (that nvidia video acceleration thingy whatsitscalled).
Didn't they show a cell decoding a lot of video streams all at once long long ago?
Peace.
Didn't they show a cell decoding a lot of video streams all at once long long ago?
Peace.
IIRC Kutaragi said they were using 2 SPEs. they were using more for SACD decoding. There ought to be quite a bit of CPU left over while playing movies.
It is very cute but here is my opportunity to state that the emperor has no clothes where it comes to F@H.
My instinct from reading is the science output of the project is minuscule. It is like Seti in this regard. The average folding at home user thinks they are cranking out new drugs, or helping a project which is cranking out new drugs. Actually what you are doing is helping a pie in the sky research project that is investigating possible maths models for protein folding simulation and testing some against lab results. Nobody is being saved with folding at home, and nobody is likely to be saved for decades. The project probably knows this but they prefer to let participants think they are "curing cancer" or some such bunk. The job of fully and accurately simulating the incredible interactions of proteins within cells is orders of magnitude more complex than the simulation and the wider issues around doing so, and using the results, are orders of magnitude more complex as well. This is really just a big phd project that has, thru the "excitement" of distributed computing (which is mostly, to be frank, stat-whoring), been gifted with a very large but inefficient and EXTREMELY power hungry compute array.
If you're interested in helping, calculate the power bill you'd save by not running F@H and then donate that money to a charity that buys vaccinations for the third world, or something like that. Most PCs go into heat/power overdrive running F@H in the "background". Leaving PS2 on vs off is over 200 watts as well. You can really make the argument that the scientific worth of folding at home is far counter-weighed by the power expense of running it.
I think these DC projects are rarely brought into sunlight and forced to justify the value of their output vs their nett power consumption (to say nothing of the hours of installing/upgrading/mucking around that goes on).
My instinct from reading is the science output of the project is minuscule. It is like Seti in this regard. The average folding at home user thinks they are cranking out new drugs, or helping a project which is cranking out new drugs. Actually what you are doing is helping a pie in the sky research project that is investigating possible maths models for protein folding simulation and testing some against lab results. Nobody is being saved with folding at home, and nobody is likely to be saved for decades. The project probably knows this but they prefer to let participants think they are "curing cancer" or some such bunk. The job of fully and accurately simulating the incredible interactions of proteins within cells is orders of magnitude more complex than the simulation and the wider issues around doing so, and using the results, are orders of magnitude more complex as well. This is really just a big phd project that has, thru the "excitement" of distributed computing (which is mostly, to be frank, stat-whoring), been gifted with a very large but inefficient and EXTREMELY power hungry compute array.
If you're interested in helping, calculate the power bill you'd save by not running F@H and then donate that money to a charity that buys vaccinations for the third world, or something like that. Most PCs go into heat/power overdrive running F@H in the "background". Leaving PS2 on vs off is over 200 watts as well. You can really make the argument that the scientific worth of folding at home is far counter-weighed by the power expense of running it.
I think these DC projects are rarely brought into sunlight and forced to justify the value of their output vs their nett power consumption (to say nothing of the hours of installing/upgrading/mucking around that goes on).
That's pretty impressive . Nice that you can browse the web while it's running too.
Stanford will have their own "more detailed announcement" on their news page later, so might be worth keeping an eye on this:
http://folding.stanford.edu/news.html
Its nice in comparison until you realize that the client optimized for ATI GPUs in PCs, can do 60 gflops..
So the entire power hungry folding collection of 155k PCs with their average 1gflop each could be replaced with 4000 PS3s, or just 2000 ATI GPUs running off the shaders .. or probably a single custom computer that used a few thousand watts.
sigh.
So the entire power hungry folding collection of 155k PCs with their average 1gflop each could be replaced with 4000 PS3s, or just 2000 ATI GPUs running off the shaders .. or probably a single custom computer that used a few thousand watts.
sigh.
Jawed
Legend
The ATI GPUs can't do all the science - some kinds of work unit can't be executed. Apparently some parts of some calculations can't be implemented. It's all a bit vague.
There was a comment that ATI's CTM technology coupled with what Stanford has learnt from the implementation of the PS3 version, could get the ATI GPUs doing a broader range of work units - but I don't know what chance there is of that happening.
I think PS3 is capable of doing all kinds of work unit (just like the PC version), making it very attractive. The other thing to note is that the GPU client is still beta, which is partly why it hasn't been taken up widely, whilst the PS3 client has completed its beta phase, so won't suffer the kind of teething problems the GPU client has had. This should mean the sheer number of PS3 folders out there ramps up quickly.
ATI's R600 will certainly be interesting: it should be 2x plus faster than X1950XTX - but it'll prolly still have a restricted set of work units it can compute.
Jawed
I think this is an entirely unfair characterization based on a strawman characterization. First of all, SETI's "science output" is a search for a counter example to the statement "there is no intelligent life anywhere else but earth within our lightcone", and as such, there is no incremental progress. It is all or nothing, and there is a very real possibility that a counter example can never be found because none exists. It's like funding "GoldbachConjecture@Home" searching for a counter example. About the only thing SETI@Home accomplished was to demonstrate that DC was feasible.
Folding is much different, and I think your characterization is asinine. You hold F@H to an impossibly high standard, that "output" consists of drugs or curing diseases. Much work in molecular biology today is basic, not applied research. Genomics, Proteomics, Transcriptomics, et al, are not worthless simply because they didn't immediately effect people's health. It's like some moron running around saying "The Human Genome Project has no clothes! It's been many years and no cure for cancer yet!" F@H has produced peer-reviewed results that have yet to be refuted.
What the average folding user thinks is irrelevent. The question is, is the CS research being done to design distributed algorithms for protein simulation of any intellectual value, and have large scale trials of the software turned up any new knowledge in computer science or biology, and I think the answer is unequivocably yes.
Another red herring. So what's you suggestion? No simulations of protein folding should be bothered with, because we can't simulate the protein inside the cytoplasm, nor can be fully run entire virtual cell simulations? The hallmark of western science is reductionism, and we can and do glean useful information by simulating or analyzing small components of massively complex systems, rather than always dealing with everything holistically.
Sure, one can't neccessarily just take the output of F@H and completely predict what a protein will do inside of a single cell, not to mention a cluster of cells. But one can't climb stairs without taking the first step, and the first step to simulating more complex protein environments is to simulate the simplest possible. As far as distributed computing goes, even if you were to assume that BlueGene petaflop scale computing became affordable by major pharamceutical companies and universities, you'd still benefit from simulation code designed for distributed processing because as you point outself, at some point, you'd probably like to network together the petaflop machines to do even more complex work.
Oh please. First of all, you could not even begin to calculate the economic benefit/scientific work of projects like this, because one can't predict scientific breakthroughs. So your purported cost-benefit argument is bunk. Scientific worth is unpredictable and serendipitious. Often pure research produces results which seemingly have zero real world applicable use, and then 20 years later, they turn out to have very important applications. R&D is essentially a search in a vast knowledge space, and that's why we must *wastefully* fund many many projects before a winner is found. It's why venture capitalists must invest in 40-100 companies on average before they turn out the unexpected breakthrough.
Secondly, the priority argument is a fallacy, as usually, it assumes all-or-nothing investment. But if you takes the tact that the $10-14/month spent running 200W 24-hr-a-day simulations would be better spent for vaccinations, you can also argue that one should not buy $3000 of HDTV equipment, or $400-600 of game consoles (4+ years of $10/mo support to the poverty stricken), or that people should drink 3 less StarBucks lattes per month, or that we money spent on Vista should go to the poor, and so on. It's a fallacy, especially for people who buy $400-600 consoles and HD setups, is that $10/month usually does not prevent them from doing other good. Or, rather than by an HDTV + console, they would have done far more good energy wise to buy a hybrid, higher MPG car, or install a solar or wind generating equipment (or purchased offsets)
And if you really want to argue the vaccine angle, the $10/mo spent on F@H work would be offset by cancelling your Xbox360 Live subscription, selling it on eBay, and buying a PS3 which has free online.
If people were forced to justify power consumption like that, they may as well ban game consoles, TiVos, air-conditioning, and force everyone to live with LED light, because people's power usage is already so so incredibly wasteful. I mean, let's get real. So many people waste hundreds of watts leaving lights on, TVs playing, wasting water, wasting gas. Someone who runs an HVAC for 1 hour will consume the power of 17.5 PS3s running F@H.
Why should anyone have to justify their computing power project to you? How much power is wasted worldwide on stupid TV shows, how much power at internet data centers is wasted on nonsense like MySpace?
The power wasted by F@H is dwarfed by the overall wastage in our economy on mindless entertainment that does not produce any scientific output and is purely consumptive. You're complaining about drops leaking out of a bucket, when water is gushing out a huge gaping hole that is Idiocracy as we know it. I think the power wasted by American Idol alone puts F@H to shame.
Really, if you want to do the most environmental or health good, you'd do well to stop wasting electricity complaining about DC projects, and instead be a blowhard electricity waster campaigning for power conservation in other areas. If you spend an hour a day convincing people to cut down on A/C, drive less, or use less hot water, than whining about PS3s running F@H, you'd be more environmentally effective.
You could also convince people to stop buying mega powerful PCs and GPUs and gaming, and instead buy power efficient internet devices that consume 1/10th the power, and do 90% of what most people need to do.
Rainbow Man
Veteran
That's a needlessly trite way of looking at things I think.
Last I checked we live in a free world. So if we choose to spend our money on having our systems run folding@home or buying vaccines for the 3rd world (which I wager - based purely on assumption and statistics of averages - you don't do all that much of anyway) OR nothing productive at all such as having our PCs or PS3s running games all day long is entirely up to us.
So I don't see where your judgemental attitude is coming from really.
At this stage distributed computing - or at least folding - can be seen as basic science research. We can't know and probably WON'T know what exactly will come out of it until soime time has passed. So it's not possible to declare its value based on nothing but appearances..
Peace.
So? And a client optimized for an R600 or G80 would score even better. Unfortunately, unlike game consoles, one can force every single F@H user to buy the most flop/watt efficient GPU. It is the mass production and homogeneous platform that guarantees PS3 F@H will dwarf the PC power contributed today. Having it installed and configured by default on platform with the good performance characteristics goes along way compared to trying to urge people to install a client and buy a good GPU.
BlueGene/L delivers 360TFlops for 1.5 Megawatt. If you assume PS3 DP is only 14GFlops, it would take 25714 PS3s to match the aggregate flops figure. Those 25714 PS3s would consume 5.1 megawatts of power, or roughly 3.4x more power. On the other hand, BlueGene/L took $140 per GFlop to build, roughly $1million per rack, and costs $50-64million depending on how you count, not counting construction and maintainence of the facillities. 25,714 PS3s would cost $20million if you assume $800 per machine.
So roughly 3.4x more power hungry, but 3x cheaper. Cheaper = more money for starving children.
Plus, unlike BG which is highly centralized, millions of people can enjoy entertainment and movies when not curing cancer and wearing emperors clothes.Human nature is to be curious and wanting to expand knowledge and put that knowledge to use. This is what folding is all about, so if we run 1 million or 2 PS3´s and fold like crazy and end up empty handed.. at least we TRIED! if we end up with some kind of medical advancement we have done good.
I hope they prioritize backround downloading before that...
On this, we're relatively sure. On the subject of single precision, though, afaicr, most folding only requires single precision math, though don't quote me on this.
And just a general reply to "onlooker1"...
This is doing valuable work, regardless of if or not anybody's produced drugs or treatments based upon it yet. The point of the work is to find possibilities, the job of people is to implement those possibilities. The first part is relatively easy, but people need time.
Now, are there computing alternatives? Yes. Are they ideal? No. GPUs are sometimes faster, but cannot process all work units - Some (the ones you're obviously referring to) are just as expensive as PS3, and they require almost as much power on their own, compared to PS3. Supercomputers are powerful enough, but they're over-designed for the job, and expensive, making them inaccessable to groups who need the processing power.
Now is it a waste of power in general? I think not. A waste of power is your business PC sitting on a desk all day long with solitaire up, not doing anything potentially productive, while your home machine sits all by its lonesome on your desk at home with minesweeper up, also not doing anything even potentially productive.
Similar threads
- Deleted member 7537
- Games Industry