SCE Joins Stanford's Folding@home Program (B3D ID=32377)

[Shifty Geezer]

Oh and this comment about FLOPs is very interesting:

unfortunately we don't know what the situation is with PS3 FLOPs, but it's very likely that every FLOP counts on PS3. It's not possible to discern from that comment what proportion of GPU calculations are done "twice".

Jawed

There's also the question of counting of conting flops which they comment on as being conservative, at 1 Flop per instruction without regard for number of operations performed. In comparisons it's fair as all machines are measured by the same amount, but it's not the same as Flops as counted in other system. ie. PS3 may be well over a Petaflop by now...

 

[DieH@rd]

PC 152 TFLOPS
PS3 598 TFLOPS
Total: 851

 

[holsty101]

unfortunately we don't know what the situation is with PS3 FLOPs, but it's very likely that every FLOP counts on PS3.

This lays it out quite well

Ivoshiee wrote:
From initial comments it seems that they expected the 10000 PS3s to achieve 1 PFLOP scale, but at the moment it is about 1/4 of that. The initial uptake has been too good to be true. Either way, it has more than doubled the FAH DC capablilities and that is good.

Those numbers were an early estimate and we've since then estimated flops more conservatively. One major issue is how to count flops. We actually do so very conservatively these days for the GPU and PS3 ports to keep these numbers from getting out of hand and in order to be able to trivially backup our statements.

The challenge with flops is how you define a flop. a+b is one flop, that's easy, but what about sqrt(a) or exp(a) or 1/sqrt(a)? There's different hardware to take care of this and we've been treating this as a single flop on the PS3 and GPU, whereas others would treat this as many (10-20) flops (eg see http://ai.stanford.edu/~paskin/slam/javadoc/javaslam/util/Flops.html#exp() ).

We've updated the FAQ to agree with this style of flop counting that we're doing now. If we were to implement something along the scheme of our link, the flops would go way up, easily by a factor of 4x or even more. However, as we start to get into the stratosphere with our total flop count, a conservative approach now will avoid naysayers in the future. We may also choose to revisit our flop counting, perhaps to make it in line with the less conservative approach taken by most other people. That would greatly increase the flop count from there.

It's surprising to see that there's a maximum of about 2 days elapsed time within which to submit a finished work unit. With work units taking about 8 hours to run, a PS3 must be folding for an average of at least around 4-5 hours per day to produce any useful results.

Yeah, I don't know what sort of commitment they're expecting from the non-dedicated Folder. Depending on % of missed W.U. they may need to start working on background functionality sooner If they can get Blu-ray playback & Folding (2-4 spu split), = a 33% reduction in Folding for 2-4 hrs, they could extend the deadlines a day, maybe half a day.

Although it looks like you can surf

Actually the web browser is locked so you cannot enter in a link. But you can go to the stats page, then click on the "Help" link on the left side of the page, and then click on the "Google" link located in the text of the page. That will open up the google surf engine and you can then enter the website through google to surf the web.

Correct? (I can't check...)

And this is nice

The world map on the PS3 is updated fairly frequently (~30 min), so Sony will be able to give a better sense of how many PS3's are running at any given moment than we can.

So, no numbers (on the map?) but quicker than the Folding@Home stats page , which is about every hour at the moment.

Quotes from Folding@Home PS3 Client.

 

[Rainbow Man]

Since I've started folding on two PS3s, I'm closing in on 100,000 points...

What about games then, done any of that?

Hehe. You're serious about folding..

Not sure if it's a sign of things to come but my apartement feels warmer with PS3 running all day long fiolding. And summer's not even here yet blaaagh..
Peace.

 

[Shifty Geezer]

We've updated the FAQ to agree with this style of flop counting that we're doing now. If we were to implement something along the scheme of our link, the flops would go way up, easily by a factor of 4x or even more. However, as we start to get into the stratosphere with our total flop count, a conservative approach now will avoid naysayers in the future. We may also choose to revisit our flop counting, perhaps to make it in line with the less conservative approach taken by most other people. That would greatly increase the flop count from there.

Not sure why they're being conservative. I mean, how can you be conservative when Flop ought to have a scientific definition? It's one atomic operation performed on a float - an ADD or a MUL or what have you. Where an instruction uses more than one of these to be calculated, even if executed in a single cycle by clever hardware, it should still count for as many atomic operations needed to perform that instruction.

 

[holsty101]

Not sure why they're being conservative. I mean, how can you be conservative when Flop ought to have a scientific definition?

I kinda thought that these parts covered that

a conservative approach now will avoid naysayers in the future .. the flops would go way up, easily by a factor of 4x or even more

Can you imagine what some of the other sites would be like already...

... that's (atleast 4x) 2.46 PFLOPS -right now-

2.46 PFOLPS!!!!1111!!! PS3 is teh R0xxorz

Nuh-uh!1! that's nvidiaFLOPS PS3 is teh SUxx0Rz!

Is not!

Is so!

I'm RIGHT!

no I'm RiGht

That's something I'd rather not think about any more

Is this a scientific definition?

The challenge with flops is how you define a flop. a+b is one flop, but what about sqrt(a) or exp(a) or 1/sqrt(a)?

Because that means nothing to me.

Then there's always

We may also choose to revisit our flop counting, perhaps to make it in line with the less conservative approach taken by most other people.

as a last resort?

 

[Shifty Geezer]

Is this a scientific definition?

No. those are caluculations you want to perform. They are executed through several atomic functions. The current flop measure is kinda like measuring the distance of a road by lengths of string. Different strings have different lengths, but you count mile long bits of string the same as metre long bits of string to keep the numbers a comfortable size. Proper measurements would place the length in pieces of string as the smallest measure of string, perhaps 1m. That'll give huge figures for lengths. Instead of roads being 100 pieces of string of assorted lengths, they'll be 10,000 pieces of uniformly sized string, and actually mean something.

The problem with suddenly massive Flop figures would (or at least ought to) be negated by the fact all figures would increase accordingly. Plus the scientific community shouldn't choose arbitary standards to appease forumites!

 

[holsty101]

The current flop measure is kinda like measuring the distance of a road by lengths of string. Different strings have different lengths, but you count mile long bits of string the same as metre long bits of string to keep the numbers a comfortable size. Proper measurements would place the length in pieces of string as the smallest measure of string , perhaps 1m. That'll give huge figures for lengths. Instead of roads being 100 pieces of string of assorted lengths, they'll be 10,000 pieces of uniformly sized string, and actually mean something.

Ugh... I think I understand. No sleep last night

I was thinking if you had, say 20m long pieces of string (any less valid than the...?) 1m long pieces of string, that ummm, the people who care about measuring other peoples string (err... peer review...? I feel like I'm trying to hard) would look at all the string on the road .. and say 'screw it. look, there's a pub'

LOL I bet I'm gonna look at this post tomorrow and go 'wtf'

Plus the scientific community shouldn't choose arbitrary standards to appease forumites!

Well yeah

 

[patsu]

PC 152 TFLOPS
PS3 598 TFLOPS
Total: 851

27260 PS3s achieving 668 TFLOPs now (Total 922 TFLOPs).

EDIT: Oops, dropped a little bto 659 TFLOPs (Total 913 TFLOPs). Perhaps people want to play game over the weekend

Now 686 TFLOPs (Total 941 TFLOPs). Go ! Go ! Go !

 

[Karma Police]

Obviously Sony has done a really cool thing here. With ATi's GPU's doing great in F@H, too, though (comparatively), why isn't MS doing the same thing for the 360?

If Sony & MS worked together they could literally cure cancer! lol..........

EDIT: Looks like I'm not the only one wondering:

http://gizmodo.com/gadgets/home-entertainment/microsoft-bring-foldinghome-to-xbox-will-ya-246748.php

 

[Karma Police]

I actually remembered MS & Sony both saying they were going to do just this back in 2005. So now Sony has done it a few months after the PS3 launch, what happened to MS doing it on the 360? Was there an unseen technical issue?

http://www.wired.com/news/technology/0,67835-0.html

 

[inefficient]

I actually remembered MS & Sony both saying they were going to do just this back in 2005. So now Sony has done it a few months after the PS3 launch, what happened to MS doing it on the 360? Was there an unseen technical issue?

http://www.wired.com/news/technology/0,67835-0.html

Unseen technical issue = red rings of death?

 

[Karma Police]

Unseen technical issue = red rings of death?

Since the 360 draws less watts than the PS3 I doubt it.

I'm afraid that MS just decided that it's not that important & gave up on the idea, which would suck big time for lots of reasons.

 

[inefficient]

Hint: The answer starts with n and ends with othing

Since the 360 draws less watts than the PS3 I doubt it.

I'm afraid that MS just decided that it's not that important & gave up on the idea, which would suck big time for lots of reasons.

What does the red ring of death have to do with watts drawn?

 

[Karma Police]

What does the red ring of death have to do with watts drawn?

Heat & failure rates. But since you were being facetious, so was I.

BTW, anyone know why the nVidia can't get their GPU's to do F@H?

EDIT: n/m here it is:

http://forum.beyond3d.com/showthread.php?t=34122

 

[Rainbow Man]

Heat & failure rates.

How hot something gerts really has no direct correlation to how muc hwattage it draws.

Obviously the hotness level depends on the mass of the object amount of cooling received and thermal resistance and loads of other stuff I don't really have a firm grasp of.

Quick and dirty example..
If you take a small piece of wire resistor and run a bit of current through it it'll likely get really hot. Even if it's just a few watrts. If you wrap it up in cotton the bundle might get so hot it catches fire.

So even though PS3 draws more watts than 360 it need not neccessarily run hotter. Or it might. One has to check to be sure.
Peace.

 

[wco81]

How are the mapping the locations of the PS3 clients?

By IP or by the personal info. we enter to get PSN accounts?

 

[Karma Police]

How hot something gerts really has no direct correlation to how muc hwattage it draws.

Obviously the hotness level depends on the mass of the object amount of cooling received and thermal resistance and loads of other stuff I don't really have a firm grasp of.

Quick and dirty example..
If you take a small piece of wire resistor and run a bit of current through it it'll likely get really hot. Even if it's just a few watrts. If you wrap it up in cotton the bundle might get so hot it catches fire.

So even though PS3 draws more watts than 360 it need not neccessarily run hotter. Or it might. One has to check to be sure.
Peace.

***Swish!!!!!!!!*** Oh how the humor continues!!

http://dictionary.reference.com/browse/facetious

 

[Shifty Geezer]

I actually remembered MS & Sony both saying they were going to do just this back in 2005.

I don't recall MS ever suggesting the use of XB360 as a DC client. Sony on the other hand have been on and on about network computing. Supposedly Folding isn't the only choice PS3 owners will get with DC.

As for the peak numbers, I wonder when PS3 owners will settle down to the 'only 10% active at any time' that we see with PCs and Linux? These PS3 owners aren't going to be running Folding forever, and eventually the TFlop will drop, like PC's is 1/10th of it's peak maximum. I'm sure we'll hit the Petaflop, and there'll be a huge press release shindig, but then performance might drop even below PCs and the actually contribution, though note-worthy, won't be the spectacle we're seeing now. Unless the PS3 Folders have a different mindset.

 

[patsu]

... but then performance might drop even below PCs and the actually contribution, though note-worthy, won't be the spectacle we're seeing now.

Yeah, that's what I figured too ...unless there is a transparent way to run F@H on PS3 without user involvement (set and forget).