PS3 will save the world

robinph

Sony are going for the moral win in the next generation as well:

PS3 to tackle world ills

iregk

Wow im touched. Totally going to swing my decision in the next gen war!!!

That Guy 901

Thats crazy! Thats alot of computing! S)

Retr0gamer

What absolute tripe and typical of sony marketing. If 10000 PS3's give you the same power as 4 of the worlds fastest supercomputers wouldn't you be better of just buying the super computers? It'll probably be cheaper since the price of the PS3 is so high. Well at least soem research group is getting a little bit of cash from sonys marketing budget and all they had to do was talk a tiny amount of ****e.

Blowfish

Retr0gamer wrote:

What absolute tripe and typical of sony marketing. If 10000 PS3's give you the same power as 4 of the worlds fastest supercomputers wouldn't you be better of just buying the super computers? It'll probably be cheaper since the price of the PS3 is so high. Well at least soem research group is getting a little bit of cash from sonys marketing budget and all they had to do was talk a tiny amount of ****e.

well the point is they don't have to buy anything. FAH is done over the net, and is designed for people who are willing to let FAH use their cpus when they aren't, in fact there is actually a boards (PC) FAH team. If 10,000 people let their ps3 be used like this (and I probably would be one of them) it would be a huge boost to the research.

[edit] plus the way the cell is designed is perfect for biological algorithms like this, and would destroy any pc cpu for these type of calculations.

LookingFor

Retr0gamer wrote:

If 10000 PS3's give you the same power as 4 of the worlds fastest supercomputers wouldn't you be better of just buying the super computers?

No, even if Stanford was buying the PS3s themselves. 10000 x $500 = $5,000,000. The BlueGene/L supercomputer that the article refers to cost the US government $100m, so 4 of them..

But as Blowfish noted above, Stanford doesn't pay a penny for the hardware - they leverage the idle time of the existing network of PS3s. But therein lies a small catch - if they bought those supercomputers, they could run their software at their convenience, all the time if they wanted (of course they'd have to pay the bills to keep them running too, though), but with PS3s, their software only runs when there's processing power to spare. So 10,000 PS3s isn't really like 4 supercomputers..they would be if they were running all the time on F@H's software, but they obviously won't be. On the other hand, they're likely to get very many more than 10k participants on PS3.

That Guy 901

Smart move on both Sony's and Stanfords part. Really doing the research some good.

Away from topic, what course would you do in college to work as this in biological research??

Retr0gamer

Either a Biotech degree or the biology option of a Common entry science degree. I'd say they have a one or two Computer based PhD students as well.

iregk

Blowfish wrote:

[edit] plus the way the cell is designed is perfect for biological algorithms like this, and would destroy any pc cpu for these type of calculations.

Right seeing as how you obviously know chip architecture inside out please give us exact specifics on what makes the cell much better for bio algorythms as opposed to say a multi core processor?

This is again Sony hopping on the marketing bull. A lot of stuff like this is already running on hope pc's, pretty much the way the Seti worked. Cancer research is now done in this fashion. What sickens me is that you never hear Dell or IBM talk about how their pc's are being used to cure cancer where as Sony have stopped so low that they have to use this as a marketing tool. Is that what the world is coming to when something like Alzheimers gets used as a marketing point?

LookingFor

iregk wrote:

Right seeing as how you obviously know chip architecture inside out please give us exact specifics on what makes the cell much better for bio algorythms as opposed to say a multi core processor?

This is again Sony hopping on the marketing bull. A lot of stuff like this is already running on hope pc's

Folding@Home is already available on the PC. The point is that PS3 will offer much higher per-unit performance than the average PC participating in Folding@Home. High end PCs with a high-end ATi graphics card will soon be able to get a client that offers comparable performance for F@H as PS3, but that market is smaller than PS3's.

iregk wrote:

What sickens me is that you never hear Dell or IBM talk about how their pc's are being used to cure cancer where as Sony have stopped so low that they have to use this as a marketing tool.

First of all, Sony hasn't said anything about this yet. The only people who've announced this is Stanford University, the Folding@Home project. Sony hasn't even made a press release for this yet.

http://folding.stanford.edu/FAQ-PS3.html

Second of all, if Intel or AMD had a commodity processor that offered 10x the performance of more typical desktop CPUs in this application, I'm sure they would trumpet it (or Folding@Home would). (And that's comparing the actual floating point performance on PS3 with F@H with peak theoretical figures on the desktop CPUs..their actual performance would be lower still).

People will be able to get similar kinds of power on the PC soon for Folding@Home using ATi's latest GPUs, but not CPUs.

Just read Stanford's own info on it at that link.

projectmayhem

basically, cell consists of one PPE and 7 functional SPEs. the PPE is dual threaded, and the SPE's are single threaded. Think of threads as "threads of execution"... PS3 will have NINE threads working simultaneously, which is a form of multi-threading. To put this in perspective, a P4 has Hyper threading, which can duplicate logic processor and get up to four threads of execution.

a thread is a "process under execution". i.e. the Pentium 60 MHz in 1994, Microsoft Word couldn't do a spell-checker until after the document was saved or when explicitly told to do so. There was a single thread of execution, which was being executed already so you could type... then after the document is finished, the same thread was executed for the spell-checker. now, in Word 2005, spell-checker works simultaneously as you type thanks to multiple threads of execution that allow MS Word to do spell-checking while you type...

so on Cell, devs can put one SPE for texture mapping, another one for sound/music, another one for physics, another one for AI, etc. Granted, this is a lot more complicated than having a single or a dual core like 360 and it will take time before coders can fully exploit the cell structure. there are other goodies in it as well, stream processing being one. there's loads of extra bits and pieces to the cell that i frankly haven't got a clue about... but there's one aspect of it for you

so basically, Cell is worth all this hype

Smellyirishman

PS3 may as well do something with it's power while it waits for a game worth playing to come out.

Seriously though, it seems like a great idea that could help vitally important research, sure Sony might spin some marketing on it but such is the nature of the beast.

Vegeta

The amount of complete bullsh1t floating around in here is crazy

projectmayhem

Vegeta wrote:

The amount of complete bullsh1t floating around in here is crazy

feel free to point it out and counter it at any time...

Vegeta

How many people in here have degrees (probably need a lot more to work in the fields being talked about) in maths? Electronic engineering? (i actually have one of these)

For people to say that a cell is much more suited to run bio algorithms than other processors, implies they must

a) understand bio algorithms, which I am sure is tough enough on its own (i think the weather is modelled on a 9 variable partial differential equation and that's just the weather not the folding of a protien)

b) have an in depth working knowledge of all processors out there and the various types of architecture designed for math applications, yeah sure they do

c)read it somewhere and believed it.

Like i know this is the playstation thread, heck i own a PS1 and a PS2 and will buy a PS3 if it is a good console but I cant stomach what some people come out with.

Blowfish

iregk wrote:

Right seeing as how you obviously know chip architecture inside out please give us exact specifics on what makes the cell much better for bio algorythms as opposed to say a multi core processor?

Well the key to a lot of biological(genome/DNA etc.), plus a lot of other scientific HPC is parallelism. If there is one thing that the cell does do well with its SPE's, its parallelism.

The performance of the cell for scientific computing has already been tested, read this, it's a research paper that came out of berkeley in may.

Overall results demonstrate the tremendous potential of the Cell architecture for scientific computations in terms of both raw performance and power efficiency

We also conclude that Cell’s heterogeneous multi-core implementation is inherently better suited to the HPC environment than homogeneous commodity multicore processors.

projectmayhem

Vegeta wrote:

How many people in here have degrees (probably need a lot more to work in the fields being talked about) in maths? Electronic engineering? (i actually have one of these)

For people to say that a cell is much more suited to run bio algorithms than other processors, implies they must

currently studying computer science, and have a keen interest in new technology. it doesn't take a degree to read up about things.

Vegeta wrote:

b) have an in depth working knowledge of all processors out there and the various types of architecture designed for math applications, yeah sure they do

i'd say my knowledge of processors isn't too bad... i know what i know. my big fat post only dealt with P4 and the old Pentium 600.. nothing magical.

c)read it somewhere and believed it.

Vegeta wrote:

I cant stomach what some people come out with.

don't read it then?

Vegeta

projectmayhem wrote:

currently studying computer science, and have a keen interest in new technology. it doesn't take a degree to read up about things.

i'd say my knowledge of processors isn't too bad... i know what i know. my big fat post only dealt with P4 and the old Pentium 600.. nothing magical.

c)read it somewhere and believed it.

don't read it then?

Well my point is that the cell while having great single-precision arithmetic it is still widely considered insufficient for scientific applications. I am highly skeptical about 10,000 networked PS3 being more faster than a custom built super computer. Yup its cheaper for stanford but when you consider the data to be number crunched will have to be written to a PS3 over an ethernet cable. Calculations performed and the data sent back over an ethernet cable. Lets assume you get the full 100mbps data rate from those. As someone with an interes in technology can you see that being quicker than a IBM or NEC computer

That paper Blowfish quoted even says this in the introduction and suggests architecture changes to make it better and more suited to scientific applications.

Sony are not stupid they put an adequate processor in their PS3. They put in a processor that would do the job. They did not put in this uber powerfull processor that is the be all and end all cos that would be a wate of money as it would not be fully utilized. you know this to be true.

Blowfish

Vegeta wrote:

Well my point is that the cell while having great single-precision arithmetic it is still widely considered insufficient for scientific applications. I am highly skeptical about 10,000 networked PS3 being more faster than a custom built super computer.

Sony are not stupid they put an adequate processor in their PS3. They put in a processor that would do the job. They did not put in this uber powerfull processor that is the be all and end all cos that would be a wate of money as it would not be fully utilized. you know this to be true.

I definitely agree with you that with 10,000 networked ps3's you aren't likely to come anywhere near bluegene/L, and that it isn't miraculously gonna be of a huge benefit to the scientific world, however its throughput for certain tasks is definitely an improvement on current pc cpu's. It also has got a fair few drawbacks though which will hamper it games wise, the lack of some things like branch prediction is a drawback, and like you said in another thread, its the software that makes the console in any case, so whether ps3 will be a success or not, wont rely completely on how powerful it is.

Vegeta

Blowfish wrote:

I definitely agree with you that with 10,000 networked ps3's you aren't likely to come anywhere near bluegene/L, and that it isn't miraculously gonna be of a huge benefit to the scientific world, however its throughput for certain tasks is definitely an improvement on current pc cpu's. It also has got a fair few drawbacks though which will hamper it games wise, the lack of some things like branch prediction is a drawback, and like you said in another thread, its the software that makes the console in any case, so whether ps3 will be a success or not, wont rely completely on how powerful it is.

If this were a discussion face to face i'd give you a good ol' handshake. Agree with you 100% there. Its a great great processor but not a reason to buy a console.

So when I hear kids going the PS3 has a "cell" they use it like a buzz word, that makes me cringe. If a cell is running sh1t software it'll still be sh1t software.

I have high hopes in the coming years for Nintendo, Ms and Sony. Not because the games look photo realistic or the processors can crunch x amount of floating point numbers blah blah blah. Its because finally the gaming companies are starting to make fun cheap games to keep us entertained between buying big titles. I love geometry wars on 360. I cant wait to see what nintendo and sony do to combat xbox live. That's what excites me.

I cant remember the last time i played a demo before xbox live. Must have been 5 or 6 years ago (i don't buy gaming mags)

projectmayhem

Vegeta wrote:

Well my point is that the cell while having great single-precision arithmetic it is still widely considered insufficient for scientific applications. I am highly skeptical about 10,000 networked PS3 being more faster than a custom built super computer. Yup its cheaper for stanford but when you consider the data to be number crunched will have to be written to a PS3 over an ethernet cable. Calculations performed and the data sent back over an ethernet cable. Lets assume you get the full 100mbps data rate from those. As someone with an interes in technology can you see that being quicker than a IBM or NEC computer

no ones saying cell is going to revolutionise the world of science, but for applications like this, it's great. it's a great piece of kit. you wont get that power in a PC for €600, that's for sure.

but you're going on now like it's as good as a Pentium4... if it was a load of sh1te it wouldn't be applied to medical or military sciences at all.

Vegeta wrote:

If a cell is running sh1t software it'll still be sh1t software.

no one will argue that point

Tar.Aldarion

Blowfish wrote:

I definitely agree with you that with 10,000 networked ps3's you aren't likely to come anywhere near bluegene/L, and that it isn't miraculously gonna be of a huge benefit to the scientific world, however its throughput for certain tasks is definitely an improvement on current pc cpu's. It also has got a fair few drawbacks though which will hamper it games wise, the lack of some things like branch prediction is a drawback, and like you said in another thread, its the software that makes the console in any case, so whether ps3 will be a success or not, wont rely completely on how powerful it is.

You doubt it why?
The researchers who actually have to do the job estimate this.
I'm sure they look at the figures in detail.
No, i won't be getting a ps3 btw.

What absolute tripe and typical of sony marketing. If 10000 PS3's give you the same power as 4 of the worlds fastest supercomputers wouldn't you be better of just buying the super computers? It'll probably be cheaper since the price of the PS3 is so high. Well at least soem research group is getting a little bit of cash from sonys marketing budget and all they had to do was talk a tiny amount of ****e.

I laughed out loud when I read this post, dude...

Blowfish

Tar.Aldarion wrote:

You doubt it why?
The researchers who actually have to do the job estimate this.
I'm sure they look at the figures in detail.
No, i won't be getting a ps3 btw.

Well usually people have a habit of using the theoretical limit, for figures like this. If it was running at the theoretical limit, it would be nothing short of miraculous. If they are using figures achieved in practice and prove me wrong however, than that can only be good news

LookingFor

Vegeta wrote:

Well my point is that the cell while having great single-precision arithmetic it is still widely considered insufficient for scientific applications.

Not really. I presume you're referring to double precision, in the case it's all relative. While its double precision performance is a lot lower than it's single precision performance, it's still "impressive" compared to other chips, to quote the authors of that paper. See page 11 of that paper with Double Precision performance comparisons - the Cell is miles ahead of the IA64 or AMD64, and with the modest changes the paper proposes, they believe it would double that speedup again.

Secondly, not all scientific applications need a lot of double precision. I doubt very much Folding@Home uses it a lot given the performance figures they're quoting (100Gflop/s on Cell).

Third, in talking about Cell's suitability for scientific applications or any applications for that matter, precision is just one issue. Others are things like the memory architecture on Cell, which lends itself very well to very high performance when leveraged properly. That's really why you see some huge speedups in some of those benchmarks, not purely the floating point capability.

Vegeta wrote:

That paper Blowfish quoted even says this in the introduction and suggests architecture changes to make it better and more suited to scientific applications.

True, but such changes would be taking it from "fast" to "really bloody fast" as their other benchmarks attest.

steviec

Wow I'm amazed by the amount of negativity here.

Before I get my 'credentials' checked up, I do have a degree in Computer Science, I have studied Computer Architecture, I've read up on the Cell out of interest, and I know nothing whatsoever about the calculations involved in biological research.

However, the people actually responsible for said biological research have said it's good so why argue with them without any real evidence? There's 100 million+ PS2s in the world. There'll probably be a similar number of PS3s. So if even a small percentage of them are running this software, and they achieve even half of the performance quoted, that can still only be a good thing surelly?

Retr0gamer

I wouldn't go believing published research papers on how fantastic the chips are. These papers are based on theory (and probably Sony marketing bull) since none of the authors would have access to the actual PS3 or Cell chip. Also just because it's a published research paper doesn't mean it's canon. As a research student I know you have to have your personal bull**** filter on when looking for papers. There are lots of journals that publish any old tat. You can also be sure that most of these 'research projects' are fund by Sony.

I notice that that paper Blowfish used is not published by any major journal and only published by the university that the project was done in. It's not exactly a high impact factor publication and very suspect.

juanjo

well I think that a more interesting point of view is to figure out how many of these "superprocessors" are gonna be available at a certain moment, connected to the network and running this software. I don't think that pple are gonna feel ok paying the electricity and broadband bills just for sony to get all the credit.

Personally I think that connectivity/availability is a much more important point in this kind of distributed software than processing power. And Ps3 is never gonna be at the same level than home/office pc's in this.

Vegeta

projectmayhem wrote:

no ones saying cell is going to revolutionise the world of science

Yes they are!! Increase the speed it takes to simulate ans visualise the folding of a protein to help cure several diseases. How is that not revolutionary

but for applications like this, it's great. it's a great piece of kit. you wont get that power in a PC for €600, that's for sure.

Consoles are sold at a loss. PCs aint.

but you're going on now like it's as good as a Pentium4... if it was a load of sh1te it wouldn't be applied to medical or military sciences at all.

In the PS3 it hasn't been used in the medical or military sciences yet. Its all talk. It hasn't been done yet.


no one will argue that point[/QUOTE]

Vegeta

steviec wrote:

However, the people actually responsible for said biological research have said it's good so why argue with them without any real evidence? There's 100 million+ PS2s in the world. There'll probably be a similar number of PS3s. So if even a small percentage of them are running this software, and they achieve even half of the performance quoted, that can still only be a good thing surelly?

That's not my arguement though. I have a problem with reporting like this before the project has even started.

I could turn around tomorrow and say 50 Wii consoles networked together will make me breakfast. I could write a paper on it too. Until they actually do it and show results of 10,000 networked PS3s beating the worlds fastest/ accurate super computer.

They system can only run as fast as its slowest component.

Going for lunch i will return to take this up

iregk

The thing is now and maybe im missing something here. If I am please correct me. My pc in the house is high spec (dual core amd 64bit) but a lot of the day it sits idle. I have it downloading and when surfing the web it aint using much resources. Its only when I turn on games that it starts to churn the processor or when I'm running my code compliers. Outside of this as I said the second core may as well not be there!

So basically if I really looked at it I'd say my box is running well below par for maybe 80% of its time. Hence why Seti rips through packets on a weekly basis.

Now lets take my xbox. The only time I ever turn that on is when I want to play games so when the box is on its running at high load. I'm going to assume that coders are looking to maximise every bit from the processor with every game in order to make it as good as it can be so in general its fair to say that my box is running at near enough capacity for 98-99% of the time!

If I got a ps3 it would be the very same. I'd only turn it on to play games! I have a much better dvd player if I want to watch a movie so when realistically would there be a lot of idle time on my ps3? My guess would be the same as there would for my xbox. None!

LookingFor

Vegeta wrote:

In the PS3 it hasn't been used in the medical or military sciences yet. Its all talk. It hasn't been done yet.

That's probably not strictly true (talking about Cell here, not PS3). Mercury Systems has shipped initial Cell systems to clients in fields from medicine to the military and more ,and expect to start shipping first orders from September. They also have more specific tailored products in the pipeline like a Cell-based powerblock for military vehicles. Others have invested significant time evaluating it and have chirped very positively about it (like defense contractor Raytheon).

iregk wrote:

If I got a ps3 it would be the very same. I'd only turn it on to play games! I have a much better dvd player if I want to watch a movie so when realistically would there be a lot of idle time on my ps3?

Well first of all I doubt your DVD player is better than a PS3, given that your PS3 will play DVDs, and HD Blu-ray movies, and upscale your DVDs to HD, and output the signal over HDMI and..so on and so on. Asides from that, I think a lot of people will, initially at least, use it for watching movies, which does leave a lot of spare processing power to soak up. F@H could run when you're doing anything except playing a game, really. Whether it suits you or not or whether your PS3 will be idle enough to make it worthwhile is an individual matter, I think.