also, each cpu has 7 times higher io bandwith... The 65Tflops estimate is for 512 8 cpu nodes, which less then earth simulators. Plus each of the cpus now has 16GFlops (and can archive it with that much memory bandwith it has), but uses less than half of the power of the old sx6. So a "earth simulator 2" with 65TFlops would have 20% less cpus and 60% less power consumption then the first.
Hah. Forgot about the darkfriends... Every 10 pages there are a few new ones... the first books that really created suspense... But after a while even the dumbest reader notices that the main chars cannot die (they are only beaten and insulted by women because they are trying to save the world), so "more darkfriends" transformed from "more dangerous enemies" to "less guys to care about in case Rands sorcery creates some more friendly fire"...
Well, not all application can run well on the massively blown up notebook that is blue gene... (low power embeded cpus, lots of density, not much memory bandwith). There will be applications that can use for example >200GB/s memory bandwith per processor...
Why? It would be more like Weapon of Mass Attention. Because if someone uses it, you are now longer forced to keep your mind on some fucking blaring box in the corner...
They expect to sell 700 during its "active lifetime", meaning until sx-9 will be out. Not all big versions, but even a 4 node one would have been HIGH in the top500 a few years ago...
I saw the news in a german forum first, and about 90% were like cool idea. Here, about 3/4 are crying like little babies who get their teeth-rotting candy taken away... Are all those prejudice about braindeath couch-potatos true in the end?
Well, f=mv^2/r. Lets assume r=0.25 cm (the centre of gravity of a blade with the 5 mm length). The speed of the centre of gravity= 1.000.000rpm/60sec/min*2*pi*2.5mm=260 m/s (about 1km/s, btw, i cannot really follow your calculation) Lets assume the blades have a total weight of 0.1g (very assumable, the thing is about 1cm*1cm*2mm, the fill factor should be less than 0.3, lets assume 10g/cm^3. That means the energy stored in the blades is 0.5 m v^2= 0.5*10^-4kg*67600m^2/s^2=3.38 joule. Less than a airgun. Inside a metal enclosure. No danger.
Errr. Put your head out of your ass, fucktard. This guy has got a doutzend "stories" the last week or 2 alone. Each of them on his blog, wich copies the original articles. Sounds very much like a hit/advertising whore.
He sold his company for 10 times more money than i will ever earn in my whole live even if my carrer goes as planned, he could have had a great live complete with his private helicopter, boat, villa, ect. The fact that this still wasnt enough for him (i.e. breaking down over his "tragic" live and becoming a drunk) earns him a position as loser, not as a hero
My first computer was a CPC6128, that had 160kbyte 3" discs... A normal divx video (or a audio cd) would use almost exactly 1 disc per second... (and to store the content of my current raid5 array, you would need a pile higher than mount everest...)
errrr. Back when doom1 was released, everyone and their dog had 14.4 or at least 9.6k modems. Which means that you spend 3 months to reduce the downloadtime about half a minute. WHOA. And nobody gave shit about some basic game at that time. That was years after shareware games like duke nukem 1&2, biohazard, commander keen 1-6 (all before doom),ect,ect. Are you sure you got the date right? (doom was 1993...)
Well, its no problem to use those good cameras as optical microsopes in makro-modus, but these target are supposed to shot from a fixed distance, plus they are BW only, so no resolution loss due to rasterization. Of course a colour or general quality target wouldnt be possible with that method, but plain resolution is... (given the fact that a laser printer can do REAL 1200dpi, which resolts to a shitload more pixels per frame then even a Canon 1Ds can resolve.
I was really amused by that story until i read the last line about the 18 year old that was killed... I guess if shit happens, its most often not a laughing matter...
GMR discs still store the data magnetically. The MR and GMR effect only replaces the normal inductance coils in the read-head. While older heads registred a voltage spike because of the magnetic flux change in the coil while the data-layer moved below the head, the new heads have a multilayer material that has a spin-sensitive resistance, so the local magnetic field created by the data on the disc spin-polarizes the electrons IN THE HEAD (nothing on the disc) and thus created a vast difference in head resistance depending on the magnetic field.
So the only difference is in how to get the MAGNETICALLY stored data back... Nothing changed in the storage per se.
These are DSPs. They can process data. Most likely phonecalls.... The giant databases that are produced that way still need some more basic data-warehousing. Like a score of IBM R6000 or so.
I read the FA. It doesnt say so exactly, but it looks like those dsp systems could do large-scale speech to text conversion. And there (with all those FFTs, ect), FPops in general and DSPs could be useful.
When i wrote this, the link was slashdotted and no mirrors present. A dsp cluster... That makes the performance even less impressive... The grape guys in japan created that much flops per asic...
Hm. 1 Tflops would place it anywhere between place 240 (if it were sustained) or 500+ (if it were peak) on the current top500. Not THAT amazing. Also, i dont quite realize how important floating point ops are in a data-warehousing application. They just pile up tons of (faxes/emails/phone recording).
Btw: Remember the story about the 5MW wind-generator a few weeks ago? The company cant sell in the us because echelon was used to sniff fax messages that were later given to a us company (kenetech windpower) which made a patent. Complete with the original tying errors. (who was it again who said "whats good for boeing is good for america"?) (story from ZEIT, titles "treason between friends", here http://hermes.zeit.de/pdf/archiv/archiv/1999/40/19 9940.nsa_2_.xml.pdf
yeah, i know. But the problem with that bays is that they are a) expensive and b) the density is so high you need noisy cooling and c) the 5.25" bays are unreachable with pata (because they are about 60cm away from the board.
But nice things, otherwise (have seen a 6 2.5" raid bay in 2 5.25", too. But same problem: too expensive....
I will go 2.5", too. The very moment i can get a 300GB 2.5" disc for under 200. Until then, 3.5" is small enough.... (my tower has still tons of unused 5.25" bays...)
Well, but you can keep your computers password (if you are always online) or the internet providers password away from them without ANY problem. And until they have reached at least 2 digit ages, this would be very advisabe (because the internet is NO childrens playground)
Slashdot Mirror
also, each cpu has 7 times higher io bandwith...
The 65Tflops estimate is for 512 8 cpu nodes, which less then earth simulators.
Plus each of the cpus now has 16GFlops (and can archive it with that much memory bandwith it has), but uses less than half of the power of the old sx6.
So a "earth simulator 2" with 65TFlops would have 20% less cpus and 60% less power consumption then the first.
Well, at least to do that he has to do SOME kind of thinking...
Hah. Forgot about the darkfriends...
Every 10 pages there are a few new ones... the first books that really created suspense... But after a while even the dumbest reader notices that the main chars cannot die (they are only beaten and insulted by women because they are trying to save the world), so "more darkfriends" transformed from "more dangerous enemies" to "less guys to care about in case Rands sorcery creates some more friendly fire"...
Well, not all application can run well on the massively blown up notebook that is blue gene...
(low power embeded cpus, lots of density, not much memory bandwith).
There will be applications that can use for example >200GB/s memory bandwith per processor...
Why? It would be more like Weapon of Mass Attention.
Because if someone uses it, you are now longer forced to keep your mind on some fucking blaring box in the corner...
They expect to sell 700 during its "active lifetime", meaning until sx-9 will be out.
Not all big versions, but even a 4 node one would have been HIGH in the top500 a few years ago...
I saw the news in a german forum first, and about 90% were like cool idea.
Here, about 3/4 are crying like little babies who get their teeth-rotting candy taken away...
Are all those prejudice about braindeath couch-potatos true in the end?
Hah. He has much to learn still.
Robert Jordan could have filled 8 pages, just for the cereal (and another 3 for they texture of those guys shoes)
Well, f=mv^2/r. /60sec/min*2*pi*2.5mm=260 m/s (about 1km/s, btw, i cannot really follow your calculation)
Lets assume r=0.25 cm (the centre of gravity of a blade with the 5 mm length).
The speed of the centre of gravity= 1.000.000rpm
Lets assume the blades have a total weight of 0.1g (very assumable, the thing is about 1cm*1cm*2mm, the fill factor should be less than 0.3, lets assume 10g/cm^3.
That means the energy stored in the blades is 0.5 m v^2= 0.5*10^-4kg*67600m^2/s^2=3.38 joule.
Less than a airgun.
Inside a metal enclosure.
No danger.
Errr. Put your head out of your ass, fucktard.
This guy has got a doutzend "stories" the last week or 2 alone. Each of them on his blog, wich copies the original articles.
Sounds very much like a hit/advertising whore.
He sold his company for 10 times more money than i will ever earn in my whole live even if my carrer goes as planned, he could have had a great live complete with his private helicopter, boat, villa, ect.
The fact that this still wasnt enough for him (i.e. breaking down over his "tragic" live and becoming a drunk) earns him a position as loser, not as a hero
My first computer was a CPC6128, that had 160kbyte 3" discs...
A normal divx video (or a audio cd) would use almost exactly 1 disc per second...
(and to store the content of my current raid5 array, you would need a pile higher than mount everest...)
errrr. Back when doom1 was released, everyone and their dog had 14.4 or at least 9.6k modems. Which means that you spend 3 months to reduce the downloadtime about half a minute. WHOA.
And nobody gave shit about some basic game at that time. That was years after shareware games like duke nukem 1&2, biohazard, commander keen 1-6 (all before doom),ect,ect.
Are you sure you got the date right? (doom was 1993...)
Well, its no problem to use those good cameras as optical microsopes in makro-modus, but these target are supposed to shot from a fixed distance, plus they are BW only, so no resolution loss due to rasterization.
Of course a colour or general quality target wouldnt be possible with that method, but plain resolution is... (given the fact that a laser printer can do REAL 1200dpi, which resolts to a shitload more pixels per frame then even a Canon 1Ds can resolve.
I was really amused by that story until i read the last line about the 18 year old that was killed...
I guess if shit happens, its most often not a laughing matter...
GMR discs still store the data magnetically.
The MR and GMR effect only replaces the normal inductance coils in the read-head. While older heads registred a voltage spike because of the magnetic flux change in the coil while the data-layer moved below the head, the new heads have a multilayer material that has a spin-sensitive resistance, so the local magnetic field created by the data on the disc spin-polarizes the electrons IN THE HEAD (nothing on the disc) and thus created a vast difference in head resistance depending on the magnetic field.
So the only difference is in how to get the MAGNETICALLY stored data back... Nothing changed in the storage per se.
These are DSPs. They can process data. Most likely phonecalls....
The giant databases that are produced that way still need some more basic data-warehousing. Like a score of IBM R6000 or so.
I read the FA.
It doesnt say so exactly, but it looks like those dsp systems could do large-scale speech to text conversion.
And there (with all those FFTs, ect), FPops in general and DSPs could be useful.
When i wrote this, the link was slashdotted and no mirrors present.
A dsp cluster... That makes the performance even less impressive...
The grape guys in japan created that much flops per asic...
Hm.
9 9940.nsa_2_.xml.pdf
1 Tflops would place it anywhere between place 240 (if it were sustained) or 500+ (if it were peak) on the current top500.
Not THAT amazing.
Also, i dont quite realize how important floating point ops are in a data-warehousing application. They just pile up tons of (faxes/emails/phone recording).
Btw: Remember the story about the 5MW wind-generator a few weeks ago?
The company cant sell in the us because echelon was used to sniff fax messages that were later given to a us company (kenetech windpower) which made a patent. Complete with the original tying errors. (who was it again who said "whats good for boeing is good for america"?)
(story from ZEIT, titles "treason between friends", here http://hermes.zeit.de/pdf/archiv/archiv/1999/40/1
Its really cool.
My personal favourite is commingtonite... A perfect example how even strict naming conventions can lead to funny results...
And its well seeded, too. I hit my bandwith cap (225kbyte/s) after 1 or 2 minutes...
yeah, i know. But the problem with that bays is that they are a) expensive and b) the density is so high you need noisy cooling and c) the 5.25" bays are unreachable with pata (because they are about 60cm away from the board.
But nice things, otherwise (have seen a 6 2.5" raid bay in 2 5.25", too. But same problem: too expensive....
I will go 2.5", too.
The very moment i can get a 300GB 2.5" disc for under 200. Until then, 3.5" is small enough....
(my tower has still tons of unused 5.25" bays...)
Well, but you can keep your computers password (if you are always online) or the internet providers password away from them without ANY problem. And until they have reached at least 2 digit ages, this would be very advisabe (because the internet is NO childrens playground)