14 replies to this topic

[Live Forever]

The Folding@Home project is getting close to crossing the 3 petaFLOP barrier overall:
http://fah-web.stanf...y?qtype=osstats

Which would be a new milestone in terms of the most computing power available to any system worldwide. (only one other supercomputer and no other distributed computing project has yet achieved 1 petaFLOP)

Edited by Live Forever, 19 August 2008 - 01:12 AM.

[sentinel]

The Folding@Home project is getting close to crossing the 3 petaFLOP barrier overall:
http://fah-web.stanf...y?qtype=osstats

Which would be a new milestone in terms of the most computing power available to any system worldwide. (only one other supercomputer and no other distributed computing project has yet achieved 1 petaFLOP)

Wow, what an incredible initiative. I wonder if they realised how succesful it would be? And it's still only the tip of the iceberg that it will become.

[RighteousReason]

what about bot nets?

[eternaltraveler]

i wish more would go into rosetta@home instead of folding@home. Rosetta does what everyone thinks folding does (protein structure prediction).

Edited by elrond, 19 August 2008 - 08:03 AM.

[RighteousReason]

what about bot nets?

e.g.

http://www.zdnet.com...39281983,00.htm

http://insidehpc.com...-supercomputer/

[sentinel]

i wish more would go into rosetta@home instead of folding@home. Rosetta does what everyone thinks folding does (protein structure prediction).

Well only 80 or so fold out of our 7000+, why not start a thread stating why RaH is better from a research/benefit/shiny prizes/Ego-massaging PPD league fanfares.

I'm not being provocative, I mean it. I had no idea what FaH was until a few weeks ago and no idea what RaH was until a few minutes ago, so go for it. Far be it for us to back the horse with the lesser multiplier.

[kismet]

i wish more would go into rosetta@home instead of folding@home. Rosetta does what everyone thinks folding does (protein structure prediction).

Could you elaborate? It'd be interesting if you could contrast the differences of those two quickly.

BTW just a minor correction of the OP, the mdgrape-3 in Japan is claimed to have reached 1 PetaFLOPS back in 2006. That supercomputer is much more noteworthy and a better comparison as neither distributed computing projects (BOINC, F@h both broke the PetaFLOPS barrier) nor mdgrape-3 can run linpack, so they are not considered supercomputers in the classical sense i.e. general purpose supercomputers.

So if we consider F@h a kind of supercomputer we have to admit there are three other "systems" that have achieved over 1 PetaFLOPS (BOINC, mdgrape-3, roadrunner).

Edited by kismet, 19 August 2008 - 12:27 PM.

[Live Forever]

what about bot nets?

Of course, there are a number of problems that I assume most here already would be aware of. (1) Only up temporarily, 2), usually destructive in nature and not used for anything productive, 3) no "official" measuring service to measure the computational power, even though the computational power is fleeting, etc, etc.)

[eternaltraveler]

Could you elaborate? It'd be interesting if you could contrast the differences of those two quickly.

wikipedia is kind enough to have done that for me


http://en.wikipedia....ki/Rosetta@Home
http://en.wikipedia....ki/Folding@home

[sentinel]

Could you elaborate? It'd be interesting if you could contrast the differences of those two quickly.

wikipedia is kind enough to have done that for me


http://en.wikipedia....ki/Rosetta@Home
http://en.wikipedia....ki/Folding@home

Hah! Remind me to hire you as head of marketing!

[Live Forever]

As of today, F@H is now above the 3 petaFLOPS barrier, mostly due to the influx of GPU clients:
http://fah-web.stanf...y?qtype=osstats
(about 3.1 petaFLOPS at the moment)


Elrond, you might find some comfort in the fact that those in the F@H project are making their research into how to squeeze the most out of those participating in their distributed computing project (through the use of GPU clients, etc) freely available to other DC projects, which will hopefully allow them to show performance increases as well, even if they keep the same amount of clients.

[dnamechanic]

As of today, F@H is now above the 3 petaFLOPS barrier, mostly due to the influx of GPU clients:
http://fah-web.stanf...y?qtype=osstats
(about 3.1 petaFLOPS at the moment)

Noticed that the 4 petaFLOP mark has been crossed.

Looks like GPUs continue to be large contributors.

Image from Stanford Client stats.

[modelcadet]

I know people aren't going to want to do this all the time, especially whilst batteries are inefficient, but...

Is there effort to make a Folding@Home Android App? iPhone App? Maybe turn on Folding@Home when you plug in to charge your cellphone...

[RighteousReason]

what about bot nets?

Of course, there are a number of problems that I assume most here already would be aware of. (1) Only up temporarily, 2), usually destructive in nature and not used for anything productive, 3) no "official" measuring service to measure the computational power, even though the computational power is fleeting, etc, etc.)

I'm just thinking, in terms of rogue AI, could be a very significant figure.

[maestro949]

I know people aren't going to want to do this all the time, especially whilst batteries are inefficient, but...

Is there effort to make a Folding@Home Android App? iPhone App? Maybe turn on Folding@Home when you plug in to charge your cellphone...

I'd be surprised to see someone invest the effort (other than to simply prove it can be done) due to their small processors and limited battery life but I wouldn't be surprised to see something emerge over the next 5-10 years. I suspect that the next gen of distributed computing apps will involve human computation as well which is perfectly suited for mobile applications.