I'm sure most of you have heard of this, but as I recently re-took up the good cause, I figured I would give it some publicity and turn it into a mini-competition

Folding@Home (F@H) is a distributed computing project that utilizes your *spare* CPU cycles to fold proteins - read more about it here: http://folding.stanford.edu/

I encourage everyone to get to folding, and to fold for team 33 ([H]ard|Folding). There is now an SMP-enabled client so those of us with dual/quad core CPUs no longer need to run multiple instances - yay! There is also an option to only fold when your laptop is running of AC to conserve battery life. Oh, and if you have an ATI video card in the X1xxx series you may be able to run it on your GPU - see http://www.stanford.edu/group/pandegroup/cgi-bin/index.php?n=English.FAQ-ATI for more info

Here is my stats page: http://fah-web.stanford.edu/cgi-bin/main.py?qtype=userpage&username=Kllrnohj and http://www.hardfolding.com/index.php?go=38&tm=33&id=196678

Now lets see who can crank out the highest PPD (Points Per Day) - I'm at 2,339 right now...

I ran folding for a while, but I didn't really like its constant memory usage (when I only had a gig of RAM on Vista and trying to game, definitely want to use as little memory as possible), so I'm running BOINC with rosetta@home (similar to folding), SETI@home, and LHC@home.

F@H doesn't use much RAM o.0 Right now I have the SMP client running and I told it to allow large WUs that "use more memory" - but its only using 55mb total...

Likewise, I haven't had any troubles gaming with it, from my underpowered school laptop to my gaming desktop rig.

I participated for a while, but I haven't lately...

One interesting thing is that if you look at the statistics, the ATI GPU's have a LOT more power than the ps3, although it might be more limited...

I never noticed that before, but it makes sense. The PS3 isn't the god of computing and power that everyone seems to think it is - it really isn't all that impressive. The Cell dominates current x86 processors at floating point, but thats about it - the cell even lacks "standards" such as branch prediction... Having said that, it isn't surprising that 7 F@H threads are getting pummeled but the massively parallel GPU with over 24+ (depending on card model, of course) math units built solely to crunch floating point numbers.

I used to do it, but my laptop ran really hot with it running, and I had it set to only run when on AC.

My school laptop holds constant at a scorching ~70C - but who cares? It is *supposed* to be designed to handle that sort of heat and load. Its stable, so I don't care.

Yeah, I'm probably gonna stick with BOINC and Seti@Home.

I had it use up to 100 MB of RAM, and with a gig of RAM on Vista, that's not trivial.

At any rate, there also appears to be a BOINC client for folding under development.

Wait a sec, Boinc doesn't let you choose to only run when in screensaver mode? That's pretty lame.

A few days ago I saw something about a scientist using a cluster of 6 or so ps3's to function as a super-computer. Apparently they are running as fast as supercomputers he used once, but he probably optimized the code to an ungodly level

I know it would be more difficult to program, but I have always wondered why more super computers aren't made up mostly of gpu's... Would it be TO hard to make them general purpose?

Actually... Take a look at http://www.cemetech.net/forum/posting.php?mode=editpost&p=70335 I remember seeing it a while ago, but apparently they have a development enviroment for it to...

So you would rather do a worthless thing like search radio frequencies for possible signs of alien life than help find cures for diseases? Good call..

@Harq: GPUs aren't really all that general purpose. I also don't know of any motherboards that let you hook up 50 PCI-E video cards As for the PS3 supercomputer - it all depends on what you define as a super computer. In pure GFLOPs the PS3 is excellent. In practical, general purpose usage, not so much.

@Harq: I believe you meant to link to http://www.cemetech.net/forum/viewtopic.php?p=70335. I don't think you really want me to edit Kllrnohj's post.

@Kllrnohj: Precisely.

Uh.. yes it does. I'm doing just that.

Errr... right... I have been having to use copy link address to post recently because Opera isn't always letting me just click that button... (That might be you though....)

I meant http://www.nvidia.com/object/tesla_computing_solutions.html

Yes, but there is a difference between being able to, and having it be fast at it. The ALUs (the part that does math and which is now being programmed for) are designed for floating point calculations, as that is what 3d game worlds rely on. A vast majority of the code you are running in your current system does not use floating point at all, and those fancy SSE/SSE2/SSE3/etc.. instruction sets go unused outside of specialized programs to take advantage of them.

Besides, this is what you were thinking of: http://developer.nvidia.com/object/cuda.html - I think I'm going to have to download it and give it a try...

The biggest problem with GPU supercomputers and the PS3 is that your program is *required* to be massively parallel. Not every problem inherently has a parallel solution, greatly limiting the effectiveness of such hardware.