Blue Gene Baby with 2TFlops built

[McFly]

http://domino.research.ibm.com/comm/bios.nsf/pages/bluegene2003.html

It's the test machine for the real blue gene supercomputer that should have 360 Teraflops.

Later machines should have 3000 Teraflops.

I got the informations from a German site: http://www.heise.de/newsticker/data/jk-14.11.03-000/

Fredi

 

[London Geezer]

wow 3 PetaFlops..... thats A LOT...

on a side note, did u guys know that in Italian, Peta sounds a lot like fart...

 

[rabidrabbit]

...that's a weird sounding fart :?

 

[London Geezer]

...that's a weird sounding fart :?

no i mean the word Peta, sounds like the word for fart (or one of...) in italian

 

[hey69]

peta puta

 

[Ender]

topic: Blue Gene baby

OMG, teh first cloned computer...

 

[Squeak]

360TFLOPS! And that’s for the test machine! That makes the Earth Simulator seem like a mere abacus (peak 40TFLOPS)!

 

[London Geezer]

think about the day (in the far far future) when we'll have things as small as our desktop being able to do that....

 

[PC-Engine]

360TFLOPS! And that’s for the test machine! That makes the Earth Simulator seem like a mere abacus (peak 40TFLOPS)!

Well it uses a lot more processors and won't be completed until 2005/2006 so it had better be more than 40TFLOPS from last year's ES

 

[bbot]

...And Sony expects to shrink that to a size of a chip by 2005. Amazing.

 

[London Geezer]

...And Sony expects to shrink that to a size of a chip by 2005. Amazing.

not all that!!! gosh.. come one let's be realistic!!!

 

[Panajev2001a]

0.5-1 TFLOPS is the target of STI CELL.

This puppy is prolly built in 90 nm and not 65 nm technology.

This puppy is not using SIMD Vector Units with a peak of 8 FP ops/cycle ( and probably a higher clock-rate ).

 

[Tagrineth]

0.5-1 TFLOPS is the target of STI CELL.

Wha? But I thought CELL was definitely going to be at least 1 TFLOP... DMGA's arguments starting to take their toll?

 

[MfA]

It would be nice to have someone officially state the TFLOPS thing again now the architecture has actually been defined and reasonable estimates can be made.

 

[ChryZ]

Wha? But I thought CELL was definitely going to be at least 1 TFLOP...

The CELL architecture for sure ( ... some day), one chip most likely not ( ... any soon). It all comes down to how much is necessary/feasable/economic. I am sure Sony will hit the right mix, at least they did with PSOne and PS2.

EDIT: mitigated a litte the impreciseness of my conjecture

 

[zidane1strife]

Wha? But I thought CELL was definitely going to be at least 1 TFLOP... DMGA's arguments starting to take their toll?

What some have said... is that ps3 might as a whole achieve 1Tflops... this is the conservative view...

But there are some.... who believe this might not be the case...

will the radicals be proven right again(psp/ps2), or will the conservative have their day this time?

Only time will tell.... if history will repeat itself again...

 

[MfA]

PSP was never much of a topic of speculation performance wise, there was the Cell in PSP business ... but if the final outcome of that was the radical view then I dont get your point.

 

[PC-Engine]

...and wouldn't it be funny if Xbox 2 also reaches 1 TFLOPS as a whole???

 

[cthellis42]

Well we still have to see if they'll have GPU performance in "ATiFLOPS" or not.

 

[SMarth]

These days FLOPS are mostly vector FMADD. It's not like the O means much else

Anyone does anything with MIPS or scalar FLOPS these days ? 8)