NC University prof created a supercomputer using eight PS3 units |
Ó
This should make every single PS3 fan proud and happy. Over at North Carolina State University, Computer Science Associate Professor Dr. Frank Mueller has built a supercomputing academic cluster using eight PlayStation 3 consoles. Being the first of its kind in the world, the set up costs around US$ 5,000 and is capable of high-performance computing. Mueller explained though that clusters is not an entirely new concept in the computing world.
Prominent businesses that use such setups include online site Google, the stock market, automotive design companies, and even scientists themselves. The professor remarked,
Scientific computing is just number crunching, which the PS3s are very good at given the Cell processor and deploying them in a cluster. Right now one limitation is the 512 megabyte RAM memory constraint, but it might be possible to retrofit more RAM. We just haven't cracked the case and explored that option yet.
Just so you guys have an idea how powerful the PS3 cluster is, take note that there is actually a list of the top 500 fastest computers in the world. The fastest is BlueGene/L, with more than 130,000 processors, at Lawrence Livermore National Laboratory.
The setup designed by Mueller doesn't break into the top 500 though. But quite interestingly, the professor added that with approximately 10,000 PS3 machines, anyone could create the fastest computer in the world. Lastly, Mueller shared that they will be exploring ways on how to fully use this setup:
Here at NC State, we will use it for educational purposes and for research. We are working with scientists to determine the needs and how our cluster can be used to their benefit, and our computer science faculty is already using the cluster to teach classes in operating systems, with parallel systems, compilers and gaming likely to follow.
Via PhysOrg
41 Jumps PSP homebrew - PSP live TV v0.3
21 Jumps PSP homebrew - PSP Live TV v0.4
17 Jumps Buy two, get one free at Best Buy
10 Jumps Atlus now mass-recruiting debuggers
Contact Us:
The QJ.net Network |
 |
| Site | Feed |
| QJ.NET | RSS |
| Nintendo DS | RSS |
| PlayStation 3 | RSS |
| PSP Updates | RSS |
| Wii | RSS |
| Xbox 360 | RSS |
| MMORPG | RSS |
| Personal Computer Games | RSS |
| iPhone - iPod Touch | RSS |
| QJ.NET Forums | RSS |
Add QJ.NET
User Favorites - December
User Favorites - December
Categories
Archives
December 2009
November 2009
October 2009
September 2009
August 2009
July 2009
June 2009
May 2009
April 2009
March 2009
February 2009
January 2009
December 2008
November 2008
October 2008
September 2008
August 2008
July 2008
June 2008
May 2008
April 2008
March 2008
February 2008
January 2008
December 2007
November 2007
October 2007
September 2007
August 2007
July 2007
June 2007
May 2007
April 2007
March 2007
February 2007
January 2007
December 2006
November 2006
October 2006
September 2006
August 2006
July 2006
June 2006
May 2006
April 2006
March 2006
February 2006
January 2006
Comments [refresh]
booo, sony will just lose money this way making other things more expensive
They have a loss at every consol sold, and i dont think buying many consols w/ no games/accessories is going to help them
ROFL PS3 sucks.. why cant they just use nornmal PCs...its much cheaper
Ooo advancement in technology these guys are just wasting time..we want games not some scientific crap
Good luck with limited Ram haha
Do you even know what a super-computer is. This setup with that much money invested in it is a technological beast when you incorporate the Cell.
This you your tipical basher rightt here:
Dan: " booo, sony will just lose money this way making other things more expensive"
but what isn't the name of the article???
"NC University prof created a supercomputer using eight PS3 units"??????
He has no idea what he's talking about. He's talking just to bash.
And WEEZY:
" ROFL PS3 sucks.. why cant they just use nornmal PCs...its much cheaper
Ooo advancement in technology these guys are just wasting time..we want games not some scientific crap"
I want the scientific crap, you don't speak for all us. Beats being stuck in the caveman era, don't ya think?!
how much of a computing performance bargain the PS3 actually is. Ironically main complaint that the researcher had was the lack of RAM that Sony built into the unit from the factory.
They should have at made the RAM upgradeable as that's the only bottle neck on the system, IMO.
Overall this is cool news and haters need to grow up.
Meh to the fanboys...
This is kewl. At least Sony has been kind enough to give us access to easily do this.
i wonder how many 360 systems would have to be combined to give similar performace
a hundred,two hundred......i wonder ;)
im playing maybe 12 to 18
For floating-point performance, about 11 or 12 360s are needed. For general purpose computing, only about 3 are needed to offer similar performance.
Damn, if you don't know anything about computers or math, don't post on a topic about computers or math. Even the latest computers, core2duo can't do the parallel processing of the cell. The x360 is way out of it's league in math processing. You think the university did this just for fun? They needed a cheap supercomputer. Unlike buying the 30-50 core2 machines it would take to equal this processing power, or the similar amount of xbox 360s (30-50 not 8, yes the xbox is similar in computing power to a modern home PC, NOT the ps3) They only had to spend $5000 on this setup. Did you notice how they could build the fastest computer in the world with a cluster of only 10K ps3's?? That is INSANE. All of the big supercomputing clusters are running a hell of a lot more processors than that! Of course, once you are looking at that many units, they will probably have ibm make some blade servers with two cell processors in each, using the new 65 nano chips... then they would only need... 2,500 of them? Maybe less? That's not even hard to do... I wonder when google will start switching over to cell processors on their big clusters.
I agree with msot of what your saying. And is the university is telling the truth about only needing 10,000 PS3's for the worlds fastest computer than thats a bargin. As the Virgin Computer with is the worlds 3rd fastest Super computer costs £320,000,000 whereas 10,000 PS3's would only be £4,000,000. I mean 4 million quid for a supercomputer! And the world fastest!
THey are just trying to show how powerful the ps3 is! and he said that there is a way to get more RAm so dont worry about that and anyways they did the same with the 360 and they needed like 18 to get the same power!
ONLY 10K ps3s?!?!
so this ghetto supercomputer (all clusters are IMO ghetto) would ONLY cost a few million? did you forget ethernet cables (idiots would use wifi), network switches, power outlets, cooling...
"The x360 is way out of it's league in math processing....the similar amount of xbox 360s (30-50 not 8, yes the xbox is similar in computing power to a modern home PC, NOT the ps3) "
You will not need anywhere near that many PCs, but I'm focussing here on the 360.
Although there are issues with Sony's OWN figures, this was how THEY compared the 360 to the PS3 for floating-point power at E3 2005.
http://www.psxextreme.com/scripts/ps3-scrs//scrbig.asp?fileID+414+scrID+4
That's right, look again!!!
http://www.psxextreme.com/scripts/ps3-scrs//scrbig.asp?fileID+414+scrID+4
SO ACCORDING TO SONY, the XBox 360 offers about HALF the floating-point power of Cell. That's 50%. So ACCORDING TO SONY, you would need about TWICE the number of 360s to match the PS3, hence that's 16 at most, not 30-50. With hyperthreading, you'll get about 50% extra performance out of each core on the 360, bringing the number of 360s needed down to about 12.
It always changes certain characters to + in replies, but doesn't do it for new comments, hence see my post below (same post, but correctly displayed links).
"The x360 is way out of it's league in math processing....the similar amount of xbox 360s (30-50 not 8, yes the xbox is similar in computing power to a modern home PC, NOT the ps3) "
You will not need anywhere near that many PCs, but I'm focussing here on the 360.
Although there are issues with Sony's OWN figures, this was how THEY compared the 360 to the PS3 for floating-point power at E3 2005.
http://www.psxextreme.com/scripts/ps3-scrs/scr.asp?scrID=4
That's right, look again!!!
http://www.psxextreme.com/scripts/ps3-scrs/scr.asp?scrID=4
SO ACCORDING TO SONY, the XBox 360 offers about HALF the floating-point power of Cell. That's 50%. So ACCORDING TO SONY, you would need about TWICE the number of 360s to match the PS3, hence that's 16 at most, not 30-50. With hyperthreading, you'll get about 50% extra performance out of each core on the 360, bringing the number of 360s needed down to about 12.
013 is the screen showing the floating-point performance of the PS3, 360, PC and Emotion Engine.
mmmm. your not being honest are you? If not, wow another thing to proove that 92percent of americans are re re's. if so, very funny
duh, that's what they've been saying about Folding on the PS3 for ages.
Wow you only need 10,000 PS3s for the world's fastest computer? Why aren't you buying it now since it's such a bargain? This means the games are really good too. Just don't read the reviews, believe what I say about how awesome motorstorm is.
Yes that 360 must be weak if you need 30-50 XBOX 360s instead of the 9 PS3s to get the same power. I did the testing since I know so much about computers and math. That's how I came up with the ultra accurate 30-50 figure. Wow the cell is in such of a league of it's own, computer manufacturers would have to be stupid to use anything else, and they're not so they're all stupid! Nobody realizes the power of the cell but me. Come on sony, show the nonbelievers what the cell is capable of. Sony? Sony? Hello? ARRRRAAHHHHH
My god you're retarded if you believe this BS.
Publicity like this will help Sony.
False...
"I did the testing since I know so much about computers and math. That's how I came up with the ultra accurate 30-50 figure."
How is not knowing whether its 30 or all the way to 50 anyway accurate? It seems like you need to do more math.
how many wii's will it take to create a supercomputer? lmao
that's a ? I would like to know too.
the reason why you need less ps3's is because the cell cpu can handle many more processes than a standard dual core cpu.. it has 7 SPEs for god's sake... the actual cell has 8 SPEs, but sony only uses 7 out of the 8... 6 are accessible for develepers while the last one is for the OS... but still... the cell is the future of processing power... i'm a proud owner of a ps3 and i support Sony and IBM... Sony because the playstation family kicks ass and IBM for the Cell microprocessor... just think how much power 10K ps3s could do... yea PCs are cool, but once again even a Core 2 Duo processor can't handle the processing strain.... also the 360 runs on a triple core processor, so don't get your hopes up... when you compare the cell and the xenon, the cell overpowers it... the cell, once again, has 7 cores, each clocked at 3.2Ghz , while the xenon has 3 core clocked at 3.2Ghz... so both are clocked at 3.2Ghz, but the cell has more cores... so the ps3 can process more data than a 360... so i'd say you need about 2.3 times more 360s to make a supercomputer of equal power to the ps3 supercomputer... that's roughly 23,000 360s... so yea... now for the supercomputer professor Mueller made, you'd need about 19 360s... now if you do the math on how much each system would cost, for the ps3 (20 gig) it would be 500X8=$4,000... now for a 360 (core system) that's 300X19=$5,700... so making a ps3-based supercomputer is much cheaper... plus it takes less systems... and once again the only downside is the 512mb RAM restriction, but they could find a way to upgrade it...
to the guy above: noob
Is the PS3 or xbox 360 better
vote on my website
Ok, mr no common sense and sunnysideup. Neither of you made a good argument as an xbox fanboy as I expected. Sunnysideup... WRONG WRONG WRONG. When you figure in the delays in processing time in splitting up the work, the xbox at half the FPCs is barely capable of 1/6 the actual workload. Sorry, but you have no damn clue, so why do you act like you do. Just because one machine can do half the FPCs, we are talking about clusters here, and supercomputing... Not playing a FPS on that piece of crap.
So 10,000 PS3s doing a divided workload would be = to 50-60K xbox 360s... Suck on it for a little while.. You don't even know how hyperthreading works, do you?
Mr fool... WTF are you here for? Motorstorm is great. Where is a bad review?? Worst I saw was a 7.9 ... That's a damn 8 out of 10... What games on your shi t system are rated an 8-10... 6 games?? Are you kidding. Out of 130+ games?? Go the hell home. You have no friends.
LOL Motostorm(30 FPS LOL) is DULL and POINTLESS, it has no feedback! a driving game with NO rumble LMAO what a pile of crap.
Forza 2 is going to crush and PS3 driving rubbish..
Go check your figures when you say there are only 6 top rated games, here is the first search google came up with -
http://uk.gamespot.com/games.html?type=top_rated&platform=1029&page_type=games
http://bluedragon.tv/
Why dont you read and mess with your own polls and give us a break will'ya?????
I would bet you 100 xbox 360's to 1 ps3, if you spent more time reading about ps3 developments rather than your own opselite 360 crab!!!!
COMMON AM GIVING YOU 100 XBOX TROLEYS RIGHT NOW TO ONE PS3 WHO EVER CAN TAKE THIS BET, COME FORTH (with proof...)
By using quantum phisics I estimate somewhere between 2 to 3 lifetime payments and 46 cents just to buy enough wiis to create a compearable cluster...
lol lol lol am just joking but I wouldnt be supprised to read about this joke of mine becoming true...
Scroll to about 3 quarters of the demo to see FORZA 2 in action, real time damage to yummmmm..
http://uk.gamespot.com/pages/video_player/popup.php?sid=6166927&pid=932731
Hey heres an idea, go learn how to spell then write your post again.
opselite ?
crab ?
COMMON AM ?
"Sunnysideup... WRONG WRONG WRONG. When you figure in the delays in processing time in splitting up the work, the xbox at half the FPCs is barely capable of 1/6 the actual workload."
LOL! Give it up kid. Pulling numbers out of your a** does not make a valid arguement. Go provide links to back up your absurd claims. What was that? You can't? Of course not, because you are completely clueless and think making up numbers equals the truth.
You can be no more than 14 or 15 years old. If you are older, then shame on you!
Still, I look forward to your reply, since it's always funny to watch the ignorant trying to look clever.
"You don't even know how hyperthreading works, do you?"
Of course I do, so why don't YOU go on to explain it to everybody here kid, I'm sure we are all interested in your knowledge on the subject. In particular, do explain some of the pros and cons of hyperthreading, and do explain it in context of the Cell processor and Xenon.
Although this was not addressed to me, I'll answer anywway;
"Go the hell home. You have no friends".
You should take your own advice kid.
Let's get back on topic please.
Ok from my point of view the 360 has 3 cores each HYPERTHREADING to simulate 6 cores, and remember from tests eariler back, it was stated that for general processing power the 360 kicked the PS3's ass.
But I shouldnt be arguing over this because its stupid really, if my arguments win im not going to win some kind of prize or gain a new friend or anything. All im going to get in the end is flaming from the other side, and while there flaming they wont even see the horrible truth that i am talking about here.
PEOPLE STOP THIS ARGUMENT, SHAKE HANDS AND MAKE UP AND GO PLAY SOME GOD DAMN GAMES.
Anyway i have 5 more achievment points to get so that i can beat the Old Spice challenge and get some cool prizes. Cya
"When you figure in the delays in processing time in splitting up the work, the xbox at half the FPCs is barely capable of 1/6 the actual workload."
1/6 the workload? oh man, just cut the bs. listen, i'm a ps3 fan not a techie but even i can tell you don't know what your talking about. all the devs say the ps3 is only slightly more powerful than the 360, but if what you said was true, then the devs would be saying the ps3 is WAY more powerful, not just slightly more. and don't talk about clusters because we dont know exactly how the ps3 are being linked, so you can't claim the 360 would be less efficient.
you are just some dumb fanboy who can't accept he was wrong. sunnysideup proved you wrong with sonys own numbers. lets see you prove your points. if you can't, then stfu, accept you were wrong and move on. please!
but this is a forum, and so people are going to discuss there points of view. sure it may get heated at times, but that's just the way it is. imo, it would be really BORING here if there were no disagreements, no arguements and everyone was nice to each other all the time.
imagine games if violence was banned! no thanks, not for me. :)
Just answer one question for everyone
Why are people working on PS3 supercomputing clusters and not 360 ones?
Simple answer will do.
Your just too easy when it comes to the obvious.
CELL.
LOL.
Actually CHUCKING..., Sunnysideup has at least been the ONLY person here to provide evidence to back up his claims in this thread, and imo, he appears to be correct. What do you have to offer other than "Ha ha False". Where's your evidence?
"Why are people working on PS3 supercomputing clusters and not 360 ones?"
How do you know people are not working on 360 supercomputing clusters?
You could have found articles about the use of Cell in clusters of course, but where are all these reports of people using PS3s in clusters before THIS report?
And yes, compare to Xenon, Cell is more powerful for single-precision floating-point calculations (unlike REAL supoercomputers where the enthasist is on double-precision arithmetic), hence it's use here.
Cell was designed to be used for numerous applications, not specifically for the PS3, unlike Xenon which is exclusive to the 360, so it makes sense that it is more likely to be used this way.
I can see a big future with the cell processor. Maybe dual core cell processors will come out for sonys later devices a bit like pentium 4 dual core.
"so i'd say you need about 2.3 times more 360s to make a supercomputer of equal power to the ps3 supercomputer..."
You're calculating 7/3, but as you correctly pointed out earlier, one of the SPEs are permanently reserved for the OS, so it's inaccessable to devs, meaning it's 6 SPEs not 7 SPEs available, therefore you'll need twice as many 360s (6/3), not 2.3 times as many (7/3).
sorry but i had to respond this post.You honestly didnt read the title on this article which proves you're just some little fanboy who wants some attention.We're sorry daddy's to busy beating mommy and are splitting up pretty soon.So you desprately seek attention and though you might go shoot up your school tomorrow and pleed insanity in court you comment could have been better.Have you not seen Gran Turismo 4 on ps3?It's not even a full game you can download it for free and you will clearly see the difference in detail weather it be vehicles or background.Even the Forza said that he hopes steps out of GT4's shade and surpass it.Now be a good little boy and grab you little mouse ,scroll back up, read very slowly(because we know how long it take you to get a grasp of things).There, you all done? Good boy!Now, grab you little mousey again, scroll down, reply on topic and support your argument.Now that wasn't so hard and for that.........here's the bullet you're gonna use to kill yourself after mommy and daddy spit.
Thanks for that link, I hadn't seen those pics from Sony's conference before.
Interesting how the fanboys haven't responded. Not content with the PS3 being a little more powerful than the 360, they have to try to convince themselves that the PS3 is several times more powerful for fp arithmetic. Sony's figures proves this is not the case, and backs up all that we've heard from developers about the PS3 compared to the 360 (better floating-point performance on the PS3, better general code performance on the 360).
well that proves that a hacked/modded system could probally run an operating system such as windows, mac, or linux depending on the processor type, i think windows requires the 86x processors or something.
Why aren't you commenting about Mister Common Sense (Not) posts in the PS3 Forum that are always bashing the PS3.
Always seem to support him and attack anyone else that contradicts him...but never have a problem with him despite his constant PS3 bashing. Why is that?
Funny how you always seem to track his conversations and back him up right away.
Now this was about how the PS3 makes a great inexpensive supercomputer cluster and somehow he had to come over onto this PS3 forum and bash the PS3
Glad to see we agree that it does make a great one and the 360 doesn't.
Isn't that what I said?
So why are you responding to me and not him?
"Why aren't you commenting about Mister Common Sense (Not) posts in the PS3 Forum that are always bashing the PS3."
Actually I do, but not as Game123. In case you haven't noticed, I'm not registered here, and therefore post under several aliases.
I also seem to recall a discussion with you were although we argued, I certainly told you what I thought of Mister Common Sense. Funny how you forgot that.
"Always seem to support him and attack anyone else that contradicts him..."
Support him? I could only see it as being viewed as support if you're a fanboy yourself, since I never bash the PS3 in my posts (and never have), but instead speak the facts as they are, not as I want them to be.
I also choose who I respond to. Hardcore fanboys with have a reputation on these forums, i.e. MCS, Vietone etc, are the kind I find a waste of time, because they are not interested in discussions, they just want to bash. If you're comparing yourself to these individuals, then fine, I'll treat you the same.
I also have no respect for those who respond to a bash by bashing themselves. The people I respect and have GOOD discussions here are those who can challenge people like MCS without bashing the 360, i.e. without going down to his level. You liar, are not such a person.
"Funny how you always seem to track his conversations and back him up right away."
And now you're being silly. If you want to see a conspiracy theories, go ahead and do so (it's actually quite amusing).
"Glad to see we agree that it does make a great one and the 360 doesn't."
And this is exactly why I respond to posts from people like you. I never said the 360 doesn't make a good 'supercomputer' setup, but as usual, you can't resist your little 360 bash. Hence as I've said already, if you can only respond to bashing with bashing of your own, don't expect me to support you.
"Of course not, because you are completely clueless and think making up numbers equals the truth." You do realize that this is all you have been doing from the start as well. You haven't provided any proof for your arguement, I don't see how you can say you have. You just act like you know something, same as the rest of the xbox trolls.
And what the... "i'm not a techie, I'm a fanboy" so you are qualified to even touch this subject? And pretend to be an expert? WTF man..
Sunnyside... Where are your links? You prove it. I don't have to be looking for online articles at crappy forums to prove what I said from experience. Unlike you, my entire education was NOT forums...
On another point... Seriously guys, are there more xbox 360 users here than PS3 users?? Are you all playing motorstorm? lol.. Since when do they run this place?!?!?! Everyone took a break.
From the READ link
"The PS3 allows the Linux operating system to be installed, and IBM designed the programming environment for programming the Cell processor (including eight vectorization units), which combined tremendous computing power within a single PS3. According to Mueller, each PS3 unit contains six operational special-purpose cores for number crunching and one general-purpose core that is two-way multithreaded in his configuration, so the eight machines clustered have 64 logical processors, providing plenty of number-crunching ability in addition to running the latest games."
"“With the PlayStation 3, you are getting the performance of a supercomputer at the price of an entry-level PC,” said Andrew Rassweiler, teardown services manager and senior analyst for iSuppli. "
"And what the... "i'm not a techie, I'm a fanboy" so you are qualified to even touch this subject? And pretend to be an expert? WTF man.. "
lol! geez you are stupid. unlike you, i didn't pretend to be a techie, that's why i SAID i'm not one, but i can tell when some kid is talking bs online, since you still have not provided any links to prove what you said.
"Sunnyside... Where are your links? You prove it."
he has you idiot. he provided the link showing sony's figures on the floating-point power between the PS3 and 360. what have you provided? he proved the 360 has half the fp power of the ps3. what have you proved? absolutely nothing!
and yet it is you who claims 30-50 360s are needed to compare to 8 ps3s. it is you who said 360 is 1/6 as powerful as a ps3 for fp. YOU made the claims, so YOU prove it or shut up boy.
"You do realize that this is all you have been doing from the start as well."
No, from the start, I provided a link to prove my point. As the guy said above, you made the claims of poor 360 performance compared to the PS3, so the enthasist is on you to provide proof of YOUR claims, since your claims go against everything devs have been saying about the two consoles and it goes against the data available.
"i'm not a techie, I'm a fanboy"
I didn't see a single fanboy comment in his post. He says he has a PS3 and not once did he bash the ps3. So, you think that anyone who dares to disagree with you is a fanboy? R-i-g-h-t.
Now this made me laugh;
"I don't have to be looking for online articles at crappy forums to prove what I said from experience. Unlike you, my entire education was NOT forums..."
Experience? What experience do you have of programming the PS3 and 360? None, so you have NO experience to use as a basis for comparing the two. And as for not looking for online articles? Oh, you mean ignore comments from experts in the field and devs working on the consoles because YOU know it ALL already. I see. LOL!
And please, don't mention education, because it's obvious you are uneducated in this particular field. BTW, I'm still waiting for your explaination of hyperthreading (in your own words of course, not cut and pasted from the internet).
Like I said, it's always funny to watch the ignorant trying to look clever, and as a result, you have been VERY entertaining. Thanks.
I don't understand this article, but it's pro PS3. As an owner of an Xbox360 I must find a way to wrap this around an insult about not having any games to play....SONY LIED....Blu-ray sucks. Sorry, fan-boy blackout. It doesn't matter I will bash this article and every article Qj does on the ps3 that doesn't have a smartass anti-sony comment. Luckily Qj's contributors do a good job of bashing the ps3 for me. The xbox is better because I bought it, nothing you say can change that.
Calm down guys!
But I hate to say it 'Mr Common Sense is a Damn Fool', although I agree with your name 100% (nice one :D), I'm sorry but I actually agree with what sunnysideup is saying here. What you've said about 360 power is WAY out, it's just completely wrong. :(
You're wasting your time right now, they want scientific crap, they don't care what you want.
lololololololol ol
Oh give it up already...
http://forum.notebookreview.com/showpost.php?p=1714904&postcount=28
http://forum.notebookreview.com/showpost.php?p=1799371&postcount=46
http://en.wikipedia.org/wiki/Cell_microprocessor
http://domino.research.ibm.com/comm/research.nsf/pages/r.arch.innovation.html
what the hell is wrong with you? No, I never said I program for the 360 or ps3... In fact, THAT IS NOT WHAT THIS ENTIRE ARGUEMENT IS ABOUT YOU JACKA$$. What part of PS3 supercomputing cluster do you not get? What part of my defending the hardware VS the 360 is not making sense to you. I do know hardware, and I do know math. What you must constantly miss in all of the articles you read is that for MATH calculations, the cell is excellent. The only comparisons about the xbox processing power is that for games, the 360 has closer power levels. When you talk about the kind of math that these supercomputers process, the processor sees a steady stream of pure floating point calculations, which is where the cell can do exponentially better.
http://folding.stanford.edu/FAQ-PS3.html folding@home is one of the most promising supercomputing/distributed computing applications to come.
You know what, I'm bored of this as well... Go play with your xbox. You have no clue what point you are even trying to make other than, ps3 suxxors, xbox ftw. whatever! lol.
As far as anyone is concerned in publicity, IBM will brings home the bacon. After all, the CELL is PowerPC technology. If Microsoft had half the brains, they'd consider dropping Intel all together and go RISC. Oh the utter chaos. Now that's something to watch...even better than wrestling.
As for the supercomputer ish, DC is a big messy web of tech that's still young. The advent of macro-computing is still in its infancy, while the idea of combining multiple processing unit computers together is a whole different level of parallel computing (or distributed computing, whichever's your slice of blueberry pie).
And yes, Core 2 DUO can't get to par with a CELL, although a Colvertown will slap the CELL silly, ONLY if you actually HAD the MONEY to BUY one. And the new experimental 80-core Intel is banking on is a juicy delight. I'm sure IBM is coming up with their own, and it would be nice to know when it bashes the Intel again to a pulp. ^_^
"After all, the CELL is PowerPC technology. If Microsoft had half the brains, they'd consider dropping Intel all together and go RISC."
Apple had PowerPC, they had it for YEARS. They evaluated Cell and found it not good enough as a desktop CPU, and as a result, they dropped PowerPC based technology in favour of Intel!
So, you were saying. ;)
No facts from at all.
"What part of my defending the hardware VS the 360 is not making sense to you. I do know hardware, and I do know math."
I also know hardware, I work in the computer industry and have done so for many years and my skills in Mathematics are excellent too. Hence I know when someone is clueless in this area. You don't even seem to appreciate the difference between single-precision and double-precision arithmetic and it's relevance to supercomputers.
In my field, for any technology I have no direct access to or experience of, I seek the word of experts (in person, via research articles or the internet) because they'll know the answers to my questions. I don't decide I know it all already and make up numbers to fit in with my opinions, which is exactly what you've done here!
"The only comparisons about the xbox processing power is that for games, the 360 has closer power levels. When you talk about the kind of math that these supercomputers process, the processor sees a steady stream of pure floating point calculations, which is where the cell can do exponentially better."
You do know that IBM desgined both processors don't you? They took their PowerPC core (PPE) and added 8 floating-point unit cores (SPEs) to create Cell. For Xenon, they took the SAME PowerPC core, enhanced the floating-point unit (from VMX32 to VMX128), and put three of these cores together to create Xenon.
Where the evidence for your claim that IBM's Xenon CPU will perform excellent in the 360 for floating-point (i.e. half of Cell or more), but would suddenly collapse to just 1/6 the power of Cell for other uses, streamed fp calculations or not.
Even if I humour you for a moment and pretend it's true, your figures of 30-50 360s (1/6 the power), are clearly numbers you pulled out of nowhere, since they have NO basis on any information you have of Xenon or VMX128 to compare with Cell. So arguing as if your figures are fact is just laughable.
"You know what, I'm bored of this as well... Go play with your xbox. You have no clue what point you are even trying to make other than, ps3 suxxors, xbox ftw. whatever! lol."
And that proves my point about you. Not once have I said I have an XBox (I have, but not a 360!), and not once did I say the PS3 sucks, since this discussion was NOT about the quality of the consoles! The fact that you think so only proves you are a fanboy who was making up numbers to make the PS3 seem more impressive than it is and the 360 less so.
But once again, thanks for giving me so much entertainment. You may be bored of me, but I found you to be a great joke. Thanks. :)
I know a lot about Cell, I've read most of those articles already.
But forum posts? Arguements from individuals online is meaningless. You can find support for ANY point of view in forums, however ridiculous it may be. They are as much proof as any posts here. Seems your education on the subject does originate from forums afterall.
So let's clarify one thing here...
The issue here is not about Cell, it's about how Cell compares to Xenon, because this is what you made wild claims about (1/6 the workload, remember?).
So go away and come back when you can provide information to back up your claims, not information about Cell or it's uses.
didn't they say that the distributed computing thing with dna sequencing was going to be released with the european ps3 launch? maybe a year ago, I don't feel like searching for it.
The reason I said I was with sunnyside before was because he backed what he said, but I've been hanging around here just WILLING you to do the same. I keep thinking you'll come back here with an awesome link and *wham*, Sunnyside pwned. I even liked your name, because Mister Common Sense deserves it, but now you've changed it for what reason? Because you lost?
Yes lost, as you failed to prove your point, which shows you did make it up like sunnyside said, and it's clear that sunnysideup knows more about the subject than you. So I'm sorry to same that the only moron I see here is you. :(
Ooo your are one of those forum guys who are constantly looking for spelling mistakes in other peoples posts so you can tell them and believe that you won a trophy dont you Mr wer...?? I do love it when trolleys like your self choose to deal with this kind of staff it only proves how far away from the subject you always are...
You must have a 360 dont you???
lollollol Dumb*****..dont post here again and go cry to your mommy.
Simply, I don't have the energy to play like this anymore. I didn't 'make it up' but from projects I've seen half the computing power does not simply mean twice the processors in the parallel processing supercomputers.
Compare top 500 computers...
Number one, the bluegene/l has 131,072 processors... and a peak of 367 trillion flops
They say that a cluster of 10,000 PS3's will be a petaflop (1000 trillion flops) machine... Only 10,000.
Now here is a comparable system. Number six on the list, but notice the huge gap in flops for the cluster.
http://www.cs.sandia.gov/platforms/Thunderbird.html
That is 9024 intel xeon 3.6 ghz processors (hyperthreading , single core). Not quite the xenon in one chip, but more power than a single core of it. That one puts out a whopping 65 trillion flops...
Break it up a little, 9024 3.6 processors / 3 = 3008 triple core 3.6 processors. That's giving the x360 Xenon a bit of a huge edge... So we up the 65 trillionflops by 3.3 to make an equal number of processors. 215 trillion flops... There is your VERY overspec'd 10,000 xbox 360 cluster, compared to the quadrillion flops of the 10,000 PS3 cluster. Oh and the real supercomputer is throwing 6gb of ram onto each dual xeon processor node. So that would be just like each xbox 360 having 9gb of ram... hmm.
Given the advantage I put on the xenons power, that is 5-1 or 6-1 on processing power!! Which is where I got all my numbers from in the first place, with a few basic calculations.
Now you can check all the nice big supercomputers at www.top500.org for yourself.
Oh here is a good comment
"Overall, they conclude that the next generation cell product needs minor hardware change to scale efficiently for double precision work, but that the first generation is already between 3 and 60 times faster, and between 10 and 200 times more power efficient, than its competitors - numbers to keep in mind when you think about Apple's triumph in arranging to get dual core Xeon CPUs from Intel for only slightly more than than four times the $89 Sony is estimated to pay for an 8+1 cell at 3.2Ghz."
And here is how to do it yourself :) I think I'll hook up the two PS3s I have someday and try this !
http://oscar.openclustergroup.org/
My last statement is simple. You don't see anyone trying to build a supercomputer out of xbox 360s DO YOU??
clovertown... :) that's all the criticism staffers get.
I can't believe that's the basis of your arguement. I'm amazed! Intel's Xeon architecture is completely different to IBM's PowerPC based Xenon CPU, and so you can draw NO conclusions of the performance of Xenon based upon a different architecture like Xeon. To do so shows a complete lack of understanding of CPU technology.
Sunny is going to laugh at you when he returns, and to be honest, I can't blame him. :|
"so you can draw NO conclusions of the performance of Xenon based upon a different architecture like Xeon."
Exactly!
I knew if I pushed hard enough he'll eventually fall on his own sword if he attempted to find proof. Perhaps he thinks Xeon is close enough because it's 'Xenon' without the 'n' in the middle. Lol! ;)
Unfortunately Player, this guy is typical of the idiots we see online all the time, where they think having a tiny bit of knowledge suddenly makes them experts.
"Sunny is going to laugh at you when he returns"
You know something, I'm done laughing at him now, instead, I pity him, I really do. :)
"My last statement is simple. You don't see anyone trying to build a supercomputer out of xbox 360s DO YOU??"
Why should they want to? Xenon is not going to be used anywhere else except the 360, but Cell is. Also, the cheapest PS3 is going to offer more floating-point performance per dollar than the cheapest 360. Xenon is stronger than Cell for general purpose code/applications, but that's not going to help the scientific community.
So I fail to see the relevance of your point. It's like someone trying to say Cell sucks because Apple rejected it, when the reason they rejected it was because Cell is not as good as other CPUs for general code and applications, hence would have been a poor choice for Apple computers.
because you don't know anything. like others said, comparing xeon to xenon is like comparing cell to athlon and making judgements on cell because of athlon's performance. it's stupid. stop pretending you know what your talking about because you dont, and everyone can see you dont. even the sony fanboys have ignored you. dumba**.
"Funny how you always seem to track his conversations and back him up right away."
Seems like you chose a name to match your actions liar. Mister Common Sense is one of the a**holes of these forums, just like Vietone and a few others. He frequently posts here but I've never seen someone called Game123 back him up 'all the time' (which is what you're saying).
I read these pages almost every day, but I've never even seen Game123 before (or at least I don't remember), which probably means he's rarely posts here (like me, I read FAR more than I post).
So what happened, you two had an arguement or something?
Ya got me, there! ^_^ I stand corrected: it's Clovertown.
"mr common sense is a damn shame on the xbox commun..." (Dude, got a shorter name? Or can I call ya MCSIADSOTXC?)
He's been forced to leave with his tail between his legs.
my old powermac g5 (2ghz ppc) just crashed while I was writing out a long answer to your ***** accusations again. Once more, you prove nothing in return. You provided one crap link that said the xbox was half as powerful as the PS3, then you assumed that they didn't know about the xbox multithreading (not hyperthreading, asswipe) so the xbox must be only a third less powerful. Wrong again. You don't see the relevance in my posts because you are so blind with fanboy hate. The ps3 is a great, powerful system and will be used for great things, like supercomputers. The xbox 360 plays a few games... lol. everything else is subpar and dated.
Then I prove that a system with half the processing power of another system does NOT only need half the processors in a supercomputing cluster, as is proven by each sucessful generation of supercomputers. But you have no retailiation to that post, other than trying to make more wild claims about my ignorance. I don't want to compare xenon to xeon, although you can't honestly believe that a single xenon chip with 512 mb of ram will outperform three xeon 3.6 processors with 3gb of ram each, can you. Well you probably can, but you believe a lot of things. You probably believe in ghosts, zombies, bigfoot, fairies, creationism, etc. Other illogical claims.
Crap just shut up about it already. I'll never believe you, I hope nobody with a brain believes you. You don't make good points, you just act like YOU know something and discredit my claims as such. Where is YOUR proof. Once more.
Simple terms. It will take a lot more than twice the amount of xbox 360's to produce the same power as a cluster of PS3s in a supercomputer. Just admit it. More than double. Just say it, yes, he is right, it will take more... Then you won't look like an ***** anymore. That is all that this was about to begin with.
Fall on my own sword? Ahh, you wish buddy.. Here's your sword, bend over and take it.
http://tinyurl.com/2hmto (This is that holy grail they said I had to prove earlier)
That's all the proof you need anymore.
Lol. I'll stick with the acronym if I can... and I sure won't leave with my tail between my legs player (97? are you 10?). Get a *****ing calculator you moron. It's not that much work to figure out. I gave you a list. All you guys do is read QJ and hope that someone posts anything pro ps3 so you can try to bash. I guess gears of war isn't that fun after all :)
**** you, you ****ing ****
Sorry... couldn't resist. I started to see too many ****s in my posts. Its like teaching little kids dealing with these guys, they never ****ing learn, and can't ****ing teach themselves, so you just have to ****ing pound it into their ****ing little brains that they are wrong and you are right !!! **** !!!!
ps qj, those are all *s... so I'm not truly cursing.
"You provided one crap link that said the xbox was half as powerful as the PS3, then you assumed that they didn't know about the xbox multithreading (not hyperthreading, asswipe) so the xbox must be only a third less powerful."
I read what he said, and he said nothing about them not knowing about hyperthreading. But as usual, you twist the facts to suit yourself.
What we've both been waiting from you is ACTUAL PROOF that Xenon in the 360 offers the performance YOU claim compared to Cell. And you haven't done so. This requires ACTUAL data of the performance of Cell compared to Xenon for realworld applications, and that data doesn't exist outside of what we see in games.
So to bring up Xeon as proof is just a joke. It doesn't matter what conclusion to arrive at with Xeon's performance, it is irrelevent when talking about Xenon. Since you can't say the two are the same, just as you can't say Xeon and Cell are the same.
"Then I prove that a system with half the processing power of another system does NOT only need half the processors in a supercomputing cluster"
Which again doesn't tell us about Xenon, because by that same arguement, a cluster built around Xenon could be significantly better, just as it is with Cell. So you should be comparing Cell to Xenon directly, not bringing in other CPUs into your weak arguement.
Sorry, but you are a waste of space. A typical online loud mouth with a small brain to match (and probably a big coward in person).
So I'm laughing at you too.
I'm at work at the moment, so I'll respond later in full. But in the meantime, here's a little thought for you.
"you can't honestly believe that a single xenon chip with 512 mb of ram will outperform three xeon 3.6 processors with 3gb of ram each, can you"
IBM designed the floating-point power within BOTH Cell and Xenon with the idea of sacrificing double-precision arithmetic to achieve massive gains in single-precision performance (compare to sp, dp on Cell is only 1/10th the performance!!!) . One VMX128 unit within the Xenon core is about equivilent to an SPE in Cell. So yes, I can 'honestly believe' that the triple core Xenon with it's sp power can do so, since Cell already does.
http://www.mc.com/literature/literature_files/Cell-Perf-Simple.pdf
Look at the graph on page 11. Cell (with 8 SPEs) is offering 15 times better performance compared to Xeon for this single-precision test. 6 SPEs should offer about 11 times the performance. Xenon, in theory, should be about half that, offering 5-6 times better performance than a single Xeon.
So like I said, you have yet to prove that Xenon offers only 1/6th the performance of Cell for sp calculations in THIS configuration, especially given that you already accepted it offers 'closer power' for games (as developers have confirmed already).
>>>"you assumed that they didn't know about the xbox multithreading (not hyperthreading, asswipe)"
Get you facts right before you insult people.
Xenon features simultaneous multithreading (SMT), which is also known as hyperthreading. Look it up.
Xenon PowerPC Processing Element details:
http://arstechnica.com/articles/paedia/cpu/xbox360-2.ars/3
"Two-way simultaneous multithreading"
http://sunsolve.sun.com/search/document.do?assetkey=1-26-101739-1
"On platforms supporting simultaneous multi-threading (Hyper-Threading technology)..."
http://arstechnica.com/articles/paedia/cpu/hyperthreading.ars
"simultaneous multithreading (SMT), also known as hyper-threading"
When you're in a hole, stop digging. :)
I'll get back to work now.
On to page two.
I mean, so what if a cluster of 360 can manage 1/2, 1/6th or 1/100th the performance of a PS3 cluster? Really, it means nothing except to those making such clusters, and it's unlikely to happen with the 360 anyway, and it means nothing for games.
That said, to add my 2 cents, I'm baffled at the Xeon comparison being made here too (referring to the guy who keeps changing his name). I would understand it if the processors were related in some way, but Xenon and Xeon? Come on.
Besides, all the benchmarks for the top supercomputers in that top500 list are for 64-bit calculations only. That's double-precision (which is needed for most scientific applications outside of visual processing), which make's the comparison to Xenon and Cell even more erroneous, since these are being used for single-precision calculations only (32-bit). As pointed out, on Cell, double-precision performance is theoretically 1/10th that of single-precision, since it's 256 GFlops peak for sp compared to only 25.6 GFlops peak for double-precision on Cell with 8 SPEs active. (Don't ask for links because the Cell specs are all over the internet).
But again, in the end, it doesn't matter, because it means nothing to the consoles.
Ok, that was more like a dollar than 2 cents, so I'll stop now and let you two continue to argue if you want to waste each other's time.
I haven't bashed the PS3 AT ALL. Neither has sunny, the only person bashing consoles in this thread is YOU. I have said nothing against the ps3 but you have against the 360.
So, just because we disagree with you, you can't take it and act like a spoilt little child. If I'm 10, then you must be 6 years old.
And funny how no-one agrees with your post. absolutely no-one. not one person has come out and defended you, but look at the ratings for each post, seems everyone rates your posts being nothing but bs. The majority speaks, and they're right. lol
"Mr Common Sense is a Damn Fool"
"Mr Common sense is a moron"
"Sunnysideup is a damn fool!!"
"sunnysideup is a damn moron"
"mrcommonsenseis adamnfoolasissu nnysideup"
"MCSIADFAISSU"
Just grow man. That's not grown up behaviour at all. That's childish. Pick a name online and have the *courage* and the maturity to stick to it, instead of changing your name all the time for childish insults.
Oh, but wait, I better be careful, or you might start calling yourself "Jez is a Damn Fool". *rolleyes*
it's not like i make an effort to hide. Everyone knows who this is :) has nothing to do with maturity or courage, damn haters.
"has nothing to do with maturity or courage, damn haters"
How is using your alias to call people names mature? Sorry, but it's childish. Can't you see that?
As for haters. What do they hate? cause the people you've been fighting haven't shown any ps3 hate from what I've seen. Unlike you who has expressed hatred towards the 360.
You're the only hater here. :|
http://www.intel.com/products/ht/hyperthreading_more.htm
"Intel calls the functionality Hyper-Threading Technology (HTT)"
Dig yourself a grave already. Just because ArsTechnica gets something wrong in terminology doesn't mean you have to believe it.
OMG there it is, you gave me the link you wanted to see oh so desperately. Shows it right there, in plain sight.
The Cell BE running 90 gflops in the single precision vs the ibm ppc 970 at 3.15 gflops... And the xbox 360 is just three of those ppc cores.
Wow, you actually showed some humility and admitted defeat.
Thanks for finally ending this!
I still don't see how you keep getting "half" out of all this. 90gflops vs 9gflops... well it's a lot more than half. That chart also showed how much more powerful a single xeon core would be over a single PPC core.
Hyper-threading is simply Intel's name for SMT. True the name belongs to Intel, but it's exactly the same thing. :|
http://www.infoworld.com/infoworld/article/02/12/06/021209plhyper_1.html
"SO ACCORDING TO SONY, the XBox 360 offers about HALF the floating-point power of Cell. That's 50%. So ACCORDING TO SONY, you would need about TWICE the number of 360s to match the PS3, hence that's 16 at most, not 30-50. With hyperthreading, you'll get about 50% extra performance out of each core on the 360, bringing the number of 360s needed down to about 12."
There is his quote, misrepresenting the calculations sony already did and adding power to the 360. You are wrong. oh I'm so worried what you think of me, really. It makes me sad.
again, you are wrong. Proving that a system with half the processing power requires exponentially higher amounts of processors to equal the same level in a cluster DOES prove that simply because the xbox 360 might have half the power of the cell DOES mean that it will take a lot more than twice as many x360s to make the same cluster!!! HOLY ***** can't you read? Can you figure ANYTHING out by yourself?? All this handholding is getting pretty damn boring. You are wrong. plain and simple. read all the facts, look through the history of all the supercomputers.
Glad to see that you get into the personal insults over the internet big man... lol, yeah right.
Um, I'm really sure that it matters what my online name is in an open forum... really... And since everyone knows who i am, why can't i keep adjusting my title to reflect my current views on who is using too much oxygen on our dying planet???
And I do hate the 360, it is inferior.
And I can't stand microsoft, and the crap products they force people to buy, calling them essential updates... So I'll fight tooth and nail till the bitter end (or when they actually do something positive)
See, not in a hole. Not digging. I'm actually standing in your driveway pissing on your car...
Intel's name. That's all I said.
I didn't say 1/6 the power, just 1/6 the usefulness. Go read it, I said at HALF the Floating point calculations, you would get 1/6th the processing in a cluster app... You asswipes can't figure anything out without just making wild bull***** up.
"The Cell BE running 90 gflops in the single precision vs the ibm ppc 970 at 3.15 gflops... And the xbox 360 is just three of those ppc cores."
The core in the 360 CPU (Xenon), is based upon IBM's PPC, just like the main core of Cell. But the difference is, IBM significantly enhanced the vector processing unit, creating vmx128 (with specs very similar to Cell's SPU).
The PPC 970 only has vmx32, which is why it's so weak for floating-point maths.
IBM explain more about Xenon here;
http://www-128.ibm.com/developerworks/power/library/pa-fpfxbox/
It's very interesting imo.
To argue about Xenon, you really should understand the processor beforehand. Saying ' the xbox 360 is just three of those ppc cores' is as far from the truth as you can get. :|
the majority doesn't want to bother talking on qj anymore because of rampant idiocy.
the three of you are not a majority. It takes all of two clicks to make a post rate no stars, so you three have it on lock. Nobody else is even reading this thread. Sunnysideup just yet again proved himself wrong, and eventually he'll figure out why i said you would need so many more xboxs... I don't see why you can't see that yourself, even with all I've given you.
If you accept it's the same, then why the correction?
He was not wrong to say hyperthreading, because that's what it is technically. I've seen other articles using the term hyper-threading instead of simultaneous multi-threading for non-Intel processors. It's lazy perhaps, and yes it belongs to Intel, but technically it's not wrong (and it's easier to say).
Especially after all of this, yes I know the xenon outperforms the ppc a little, and has three of the ppe cores that runs the cell, why is everyone disagreeing with little points STILL.. Who cares. I'm making a much bigger statement here that nobody still catches. Half of the processing power will end up being about 1/6 the processors needed.
if they are going to correct every little possible flaw I might write, then I'm sure as hell going to correct them. hyperthreading can also be called smt, but because it's the intel term for it, in intel processors, it's not correct to call smt as hyperthreading.
Besides, sunnysideup has more mail and coffee to deliver around the office.
"Especially after all of this, yes I know the xenon outperforms the ppc a little, and has three of the ppe cores that runs the cell"
It doesn't outperform the ppc a little, it outperforms it a lot, that's why it's offering such great performance in the 360. You only weaken your arguement by such comments because you clearly do not want to give credit where it's due. IBM did a great job creating Xenon, just as they did with Cell. Just because you hate a company and it's products, doesn't mean you have the blind yourself to the facts.
Also, the PPE core in Cell features vmx32, not the enhanced vmx128, so Xenon doesn't have three of the PPE cores seen in Cell.
"I'm making a much bigger statement here that nobody still catches. Half of the processing power will end up being about 1/6 the processors needed."
Perhaps if you stop making false statements about the 360 and Xenon (such as those about the ppc core), then your arguement about clusters maybe heard?
You may be right on the cluster issue (I haven't checked it out myself), but if you keep making false statements and deliberately play down the technology in the 360, then all you're doing is distracting everyone from the arguement you trying to make by coming over as a fanboy. In otherwords, your message is being lost.
Instead, offer a clear and concise explaination of the issues with clusters and then perhaps people will listen.
In particular, can you show that in almost ALL cases (not just one example) CPU A with half the processing of power of CPU B will only produce 1/6th the total processing power for a given cluster size! That's what you should be showing here. Only then will your point will be proved. And what about other factors such as the size of the cluster. How does it effect performance (i.e. is it always 1/6th power despite cluster size, if not, how does it change with size). Finally, support your statements with links.
So, can you do this? Can you put down the knife for a moment and present such as arguement? If so, great, I would be interested to read it. But if it's just an exercise to bash MS/360/Xenon etc, then it would be pointless.
I hate it when I read my posts and they are full of typos. :)
Anyway, just in case my message is lost here, I'll post it again as a new post below.
(Repeat of part of my post above)...
You may be right on the cluster issue (I haven't checked it out myself), but if you keep making false statements and deliberately play down the technology in the 360, then all you're doing is distracting everyone from the arguement you trying to make by coming over as a fanboy. In otherwords, your message is being lost.
Instead, please offer a clear and concise explaination of the issues with clusters and then perhaps people will listen.
In particular, can you show that in almost ALL cases (not just one example) CPU A with half the processing of power of CPU B will only produce 1/6th the total processing power for a given cluster size! That's what you should be showing here. Only then will your point be proved. And what about other factors such as the size of the cluster. How does it effect performance (i.e. is it always 1/6th power despite cluster size, if not, how does it change with cluster size). Finally, support your statements with links.
So, can you do this? Can you put down the knife for a moment and present such an arguement? If so, great, I would be interested to read it. But if it's just an exercise to bash MS/360/Xenon etc, then forget it, since it would be pointless.
"I mean, so what if a cluster of 360 can manage 1/2, 1/6th or 1/100th the performance of a PS3 cluster? Really, it means nothing except to those making such clusters, and it's unlikely to happen with the 360 anyway, and it means nothing for games."
I agree, it is meaningless because it's unlikely to happen, however I am curious about CPU power and cluster performance, hence my request to mcsiadfaissu to expand on the point he was making (since it had been lost in fight).
for these numbers.
like, back in time.
2048 xeon 3.6 myrinet max 10650 gflops 2.95 per mhz
1024 xeon 1.8 myrinet max 2207 gflops (4414 gflops, 2048 processors) 2.45 per mhz
Half the power, less than half the performance
2048 PPC 970, 2.5 GHz max 15040 gflops 6.01 per mhz
10240 PPC 970, 2.3 GHz max 62630 gflops (12526 gflops, 2048 processors)
5.44 per mhz
512 pSeries 690 1.1GHz max 1170 gflops (4680 gflops, 2048 processors) 4.25 per mhz
Slower the processors get, the less efficient they get. Can't beat the advantage of having more of the processors on the same chip.
"Half the power, less than half the performance"
True, but it's 0.41 instead of 0.5, which is still good considering we're talking about thousands of processors in the clusters.
"10240 PPC 970, 2.3 GHz max 62630 gflops (12526 gflops, 2048 processors)
5.44 per mhz"
You haven't provided the link , but isn't that the MareNostrum - BladeCenter JS21 Cluster?
Anyway, big oops, your figure is wrong. :(
It should be 125260 GFlops for 2048 processors (62630*2). Which is close to what you would expect (4 times the GFlop performance for 5 times the number of processors!).
The pSeries 690 cluster is difficult to compare here without info on how this processor compares to the PPC 970 for floating-point ops. But even then, compared to 2048 PPC 970s, it's 1/4 the no. of processors and less than half the clock speed. If the fp power was the same, the numbers suggest 1/9th the performance, but the reality is 1/13th. Not far off (but still questionable without fp performance figures of the CPUs).
Therefore none of these clusters demonstrate the massive 1/6th performance difference you quoted earlier for a CPU with half the floating-point performance in a cluster. There's a small difference for such large clusters, but for 8 consoles?
And that's the other point, we're talking about thousands of processors in these clusters. For the PS3, we're talking about a cluster of just 8 machines! The difference seen on a mere 8 or so consoles is going to be negligable. Hence if we choose to use only 3 SPEs in each PS3 (i.e. half the power), I just can't see how we'd need 30-50 PS3s to be comparible to a cluster of 8 PS3s with 6 SPEs in use, especially given the figures you've provided for those huge clusters. It just doesn't add up. :|
BTW, how did you arrive at that figure of 1/6th? I can't see it in these numbers.
(Apologies in advance for any typos)
"Actually I do, but not as Game123. In case you haven't noticed, I'm not registered here, and therefore post under several aliases."
And I would know that how?
I've posted in response to Mister Common Sense several times followed by Game123 responding to me not him. Seems like a reasonable conclusion.
"I also seem to recall a discussion with you were although we argued, I certainly told you what I thought of Mister Common Sense. Funny how you forgot that."
Actually I did..see Subject.
"Support him?"
See First point.
"I also choose who I respond to. Hardcore fanboys with have a reputation on these forums, i.e. MCS, Vietone etc, are the kind I find a waste of time, because they are not interested in discussions, they just want to bash. If you're comparing yourself to these individuals, then fine, I'll treat you the same"
At want point did I compare myself to them? Who's bashing now?
"I also have no respect for those who respond to a bash by bashing themselves."
See Previous point.
"The people I respect and have GOOD discussions here are those who can challenge people like MCS without bashing the 360,
At want point did I bash the 360 as you like to characterize it? PS3 was designed and built to allow an open platform like Linux to be installed. 360 was built to be a closed proprietary system that wouldn't. I don't need to check. Understanding basic principles saves a lot of leg work. One facilitates exactly what this article was talking about and the other one doesn't. That's why the 360 doesn't make a good supercomputer. Please explain how stating the obvious is bashing the 360?.
" i.e. without going down to his level. You liar, are not such a person."
So I don't actually bash the 360 (ie. a console) - just state the obvious - but you bash me (ie. a person). Who's bashing and going down to what level?
"Funny how you always seem to track his conversations and back him up right away."
See Point 1
" (it's actually quite amusing)."
Bashing? Just couldn't resist your little "liar" bash? See point 1.
"Hence as I've said already, if you can only respond to bashing with bashing of your own,"
See Previous point
" don't expect me to support you. "
Exactly.
And support what exactly?
"Seems like you chose a name to match your actions liar."
I know ironic isn't it.
"Mister Common Sense is one of the a**holes of these forums, just like Vietone and a few others."
Actually they are a bit more than they seem.
" He frequently posts here but I've never seen someone called Game123 back him up 'all the time' (which is what you're saying).
I read these pages almost every day, but I've never even seen Game123 before (or at least I don't remember), which probably means he's rarely posts here (like me, I read FAR more than I post)."
See my Points in previous response (old age and busy life) to Game123. That will make it clearer.
all of the links are on top500.org
Your math is wrong. That is 10 thousand 240 processors at 62630 gflops, which is about 12526 gflops for only 2048 processors. Sorry ;)
I compared the pseries 690 cluster to the ppc 970 because it's the last generation of the powerpc architecture. They were the power4 series, and the 970 is the power 5.
And it was 30-50 xbox 360s to compare to 8 ps3s... not apples to apples... So my initial statement of 30 might not be that far off... Ok, 25. Calculated advantage from the history of ppc in supercomputers, 6.01/4.25 advantage multiplier of 1.41. 8 ps3s, 54 logical processors. * 1.41 = 76 processors /3 (xenon) 25. So 25. Probably more, but 25 is a fair starting point. Until we actually see someone try it, none of this matters. I'm not the one who came here and said, oh you only need 12 xbox 360s to equal 8 ps3s...There was NO math or truth involved in my sunnysideups original statements. Mine were hypothetical, but based in reality. Not blatant guesses like they want you to believe. It's all a moot point, as you would need 100,000,000 xbox 360's with 100,000,000 monkeys entering numbers to equal the supercomputing power of the 8 ps3s, since you don't get any access to that processor. I don't even know how this was all started!
If distributed computing gets popular, and we can get a majority of online ps3 users to allow the network to perform calculation work units on their machines when they aren't busy using it, the world may be changed... Can't say that about a game console. :) Imagine five - ten - twenty million ps3s distributed online, when the fastest computer in the world only needs 10000...
"Your math is wrong. That is 10 thousand 240 processors at 62630 gflops, which is about 12526 gflops for only 2048 processors. Sorry ;)"
Ok, my error here, thanks, but the result is exactly the same.
You say it's 12526 GFlops for 2048 processors. Correct. So it should also be 'about' 13615 GFlops for 2048 2.5GHz processors, which compared to the 2048 cluster is 0.9 times the power obtained. Again, VERY close (13615 vs 15615).
So these large clusters don't show the big difference you claim.
"6.01/4.25 advantage multiplier of 1.41. 8 ps3s, 54 logical processors. * 1.41 + 76 processors /3 (xenon) 25. So 25. Probably more, but 25 is a fair starting point."
Given your numbers, you're saying PPC 970 compared to pSeries 670 represents Cell compared to Xenon, but I don't think there's any evidence for that. Thus you say in a cluster, each SPE core in the PS3 will offer 40% more performance than each core in Xenon based on numbers from completely unrelated processors. So right now, we have to assume the performance is about the same, not 40% in favour of SPE. That's the starting point. The cluster information you provided shows very small efficiency differences for huge clusters, and I would expect to see improved efficiency as the cluster size decreases (i.e. fewer nodes).
The other assumption I'm seeing in this thread is in Sunnysideup's figures, where he seems to be assuming only 6 SPEs will be available per PS3. Well, I'm not sure of the exact configuration here, but I'm sure it's not running under the PS3 OS with one SPE reserved for the OS, so I can only assume 7 SPEs will be available per PS3. Hence by his figures, each PS3 is not 50% more powerful with SMT and 100% more without, but 75% and 133%. Hence it should be 14 360s at best (with SMT) and 18 or 19 360s without SMT.
Take out your 1.41 figure (because it's not valid here), and it's 56 arithmetic processors (8*7 SPEs) divided by 3 vmx128 per Xenon, which is 7*8/3 or 18-19 XBoxs. Hmmm, again without SMT.
So imo, you are both wrong! ;)
But with SMT, I think Sunny's figures are far closer than your 30-50 (and now 25) estimate. His error was to assume 6 SPEs instead of 7 SPEs, your error was to assume a massive difference due to having half the power in a *tiny* cluster.
Sorry guys, but that's how I see it.
...Jez for bringing calm to this discussion. Great posts. In just a few posts you've gained my respect. So I will put down my knife too. ;)
I'll start by saying sorry to MCSIS... (for the insults that is), and hopefully he will be willing to do the same. Agreed?
Anyway...
"The other assumption I'm seeing in this thread is in Sunnysideup's figures, where he seems to be assuming only 6 SPEs will be available per PS3."
Hmmm, I was so busy focussing on the 30-50 and 1/6th claim that I didn't bother to look back. You are right, I was assuming 6 SPEs, it should be 7 like you said (my bad). But hey, we all make mistakes (as you did yourself above ;)). And nit-picking, it's the 690 not 670. ;)
So I agree, it's 14 * 360s, not 12 like I said.
"you say in a cluster, each SPE core in the PS3 will offer 40% more performance than each core in Xenon based on numbers from completely unrelated processors."
Which you're right to question. I looked it up on the net, and the PPC 970 is based on the Power5 (G5) and the 690 on the Power4/Power4+.
As mentioned here;
http://everythingapple.blogspot.com/2004_07_11_everythingapple_archive.html
"SPECfp2000
POWER4+ @ 1.7 GHz - 1642
POWER5 @ 1.9 GHz - 2576
...Clock for clock, POWER5 is 40% faster than POWER4+."
The difference closely matches that seen in the cluster performance. So this is not representative of one SPE compared to one VMX128, which MCSIS uses in his calculations, this is representative of the difference between the fp power of the CPUs in the clusters. Therefore given the floating-point difference, you could have closely predicted the performance of the 690 cluster compared to the 970.
"But with SMT, I think Sunny's figures are far closer than your 30-50 (and now 25) estimate. His error was to assume 6 SPEs instead of 7 SPEs, your error was to assume a massive difference due to having half the power in a *tiny* cluster."
Agreed, and cnce again. Thanks
"If distributed computing gets popular, and we can get a majority of online ps3 users to allow the network to perform calculation work units on their machines when they aren't busy using it, the world may be changed... "
But internet speeds will not allow it to have any real practical uses, other than for the kind of things we've seen already on the PC, such as the screen saver to process calculations for the Alien search project (or whatever it's called :)).
an open platform that's ideal for supercomputing as well.
http://www.gametrailers.com/player.php?id+17786+type+wmv
Looks like you don't need the expense of another pc either. Pretty impressive stuff.
"I mean, so what if a cluster of 360 can manage 1/2, 1/6th or 1/100th the performance of a PS3 cluster? Really, it means nothing except to those making such clusters, and it's unlikely to happen with the 360 anyway, and it means nothing for games."
Good point. Still it was fun while it lasted.
It's all quiet now...
*tumble weed... wind blowing... more tumble weed...*
lol. I got a cold today and I slept in... Oh so sorry :)
Yeah whatever, we all apologize. Sorry I yelled. Supercomputers are cool.
Internet speeds have nothing to do with that sunnysideup.
"Observational Data is recorded on 35 Gigabyte tapes at the Arecibo Observatory in Puerto Rico, each holding 15.5 hours of observations, which are then mailed to Berkeley. Once there, it is divided in both time and frequency domains work units of 107 seconds of data, or approximately 0.35 MB, which overlap in time but not in frequency. These work units then get sent from the SETI@home server over the Internet to people around the world to analyze."
They send you a small part, your computer does the math, and sends back the result. If your computer is a PS3, it does 8x the math (don't start this again) and returns the result 8x faster.
I had seti@home running from 2000-2005 or so. The new BOINC setup is much smoother. Everyone should have the boinc client on their computer.
Thanks, I was trying to remember that... SETI - Search for Extra-Terrestrial Intelligence. I just noticed that mine (Alien Search Project) would have been ASP, which is not quite the right name to use for a 'friendly' search for Alien life. :)
"They send you a small part, your computer does the math, and sends back the result. If your computer is a PS3, it does 8x the math (don't start this again) and returns the result 8x faster."
Agreed in principle, that's why I mentioned it as the kind of application it's suitable for. My point about the internet was a bit clumsy, I meant that it is too slow to allow for the kind of realtime applications which is possible to achieve over a LAN (I know you didn't imply that). But in future, as internet speeds increase for users, connecting consoles/PCs in this way for uses other than SETI type apps could become a reality, but by then we would have moved well beyond this current gen of consoles.
"I had seti@home running from 2000-2005 or so. The new BOINC setup is much smoother. Everyone should have the boinc client on their computer."
I had it too about 5 years ago, but didn't keep it that long (6 months maybe). I'll check out BOINC, thanks.
Anyway, onto _that_ other issue, I think you'll agree we have to agree to disagree. Agreed? :)
Bye
(Hope you fully recover from your cold soon)
First, sorry to hear you have a cold. Get well soon. :)
I'm please you guys mentioned SETI, because it fits in with my understanding of how clusters are used.
For my understanding, the more data sent back and forth between nodes, the less efficient the cluster will run. Hence the aim is always to give each node as much data as it needs to perform a task, leave it to process that data, and then wait for the results. In other words, tell each node "Here's your data, this is what I want you to do with it, now get on with it and give me the result when you're finished". Of course, the task has to be one which lends itself to parallelization .
So taking that cluster of 8 PS3s. If a scientific computing task takes one PS3 8 minutes to complete, then if that task can be split into 8 equal sized smaller tasks, then you could give each smaller task to a PS3 in the cluster together with the data it needs, where it will process that data and return the result about 1 minute later. Finally, the results can be combined by one PS3 to produce the end result. Hence you get around 8x improvement in performance.
If it is correct that the 360 offers half the processing performance of the PS3, then to get similar performance, the same task could be split into 16 smaller tasks and spread over 16 360s (nodes). Again, it would take 1 minute for each 360 to complete it's task, and then all the results can be combined by one of the 360s for the end result.
Another example is realtime 3D, where you could split the screen into sections. Each PS3 can be loaded with all the data for the 3D world at the start, then during animation, only process the polygons and pixels in their area of the screen).
N times per second, each PS3 will send it's completed section of the image to a 'master' PS3 which will arrange them into one image. Again, the bulk of the work is being done on each PS3 processing the 3D image, minimising data being sent between nodes.
For the 360, the screen will need to be split into double the number of sections (i.e. each half the size of the PS3 sections), which again will be processed by 360 nodes and sent to one 360 for displaying as a combined final image (N times per second).
The performance difference between the PS3 and 360 clusters in the examples I've given would be negligible.
Of course, like I said, this is assuming the fp performance of the 360 is half that of Cell.
Anyway, sorry to waffle on about this, just expressing my final thoughts.
Thanks for the discussion. :)
I've just read the original article related to this topic;
http://www.physorg.com/news92674403.html
Dr. Frank Mueller, talking about his cluster, says "each PS3 unit contains SIX operational special-purpose cores for number crunching...". It seems only 6 SPEs are available per PS3 afterall, so your statement was right about the number of SPEs for this cluster of PS3s networked together.
So I'll go back to your figure of 12 x 360s with SMT (16 x 360s without SMT), all assuming Xenon fp performance is half that of Cell.
I'll leave now ...
*(puff of smoke) + (fanfare)*
four5769
and2070
OS Type Current TFLOPS* Active CPUs Total CPUs
Windows 152TFLOPS w/ 159,858 active
PLAYSTATION®3 615TFLOPS with 25,082 active CPUs
Total 869TFLOPS with 222,428 active CPUs
Total number of non-Anonymous donators = 627448
Last updated at Sat, 24 Mar 2007 11:50:43
Conveniently left out the rest of the sentence...
"According to Mueller, each PS3 unit contains six operational special-purpose cores for number crunching and one general-purpose core that is two-way multithreaded in his configuration, so the eight machines clustered have 64 logical processors."