StarDestroyer.Net BBS

http://www.msnbc.msn.com/id/42238358/ns ... d_gadgets/

‎"One petaflop is equal to 1,000 teraflops, which is in turn equal to 1,000 billion floating point operations per second. It has taken a long time to reach this much-sought goal. The first commercial supercomputer to break through the 1 ter...aflops barrier was arguably--and I say arguably because people argue about this--was probably the Cray T3E-900, which had 2,048 processors when it shipped in November 1996 and was rated at 1.8 teraflops."

That is the best information available that describes what exactly a petaflop is. There is no "Conversion Chart" available. Check out the article below it goes into more detail about the nature of petaflops.

One last piece of info.
A petaflop is measured by Floating points.
Floating-point is considered to be a method of encoding real numbers within the limits of finite precision available on computers.
Now we have the question what exactly is a floating point?
The answer; A floating-point number is expressed as a basic number, an exponent, and a number base which is usually ten but may also be 2.

SO WHAT THIS ALL MEANS IS.....
In layman's terms the best way I can describe the speed of a petaflop to you is; A petaflops is equal to 10 times the speed of all the networked computers in the US combined!!!

Source(s):

http://www.itjungle.com/tlb/tlb061708-story01.html

A flop is not a floating point, flop stands for FLoating point OPeration - division, addition, subtraction, shifts, whatever. I'm not aware of any major computing platform that uses a base of 10 rather than 2 for its floating point specifications, but if you're going to say 'usually 10' then feel free to enlighten us.

Considering that individual graphics cards are in the multi-teraflop range, your claim that

Collossus wrote:A petaflops is equal to 10 times the speed of all the networked computers in the US combined!!!

Is pretty laughable. Above and beyond 'A petaflops'.

Collossus wrote:http://www.msnbc.msn.com/id/42238358/ns ... d_gadgets/

‎"One petaflop is equal to 1,000 teraflops, which is in turn equal to 1,000 billion floating point operations per second. It has taken a long time to reach this much-sought goal. The first commercial supercomputer to break through the 1 ter...aflops barrier was arguably--and I say arguably because people argue about this--was probably the Cray T3E-900, which had 2,048 processors when it shipped in November 1996 and was rated at 1.8 teraflops."

I didn't know where else to put it, so I will stick it here: these High Performance Computers aren't just larger CPUs, they are a network of many many MANY computers. Very often they will not tackle a single problem with all available computing power, instead you get say 300 cores for 4 hours and the rest is aloted to other applications in similar chunks. Thus the FLOPS alone don't tell you much about how long your massively complex simulation, SATsolver etc. is going to have to run.

Collossus wrote:That is the best information available that describes what exactly a petaflop is. There is no "Conversion Chart" available. Check out the article below it goes into more detail about the nature of petaflops.

Just read this: http://en.wikipedia.org/wiki/FLOPS Also, what kind of "Conversion Chart" do you want? I.e. converting from what and into what?

Collossus wrote:One last piece of info.
A petaflop is measured by Floating points.
Floating-point is considered to be a method of encoding real numbers within the limits of finite precision available on computers.
Now we have the question what exactly is a floating point?
The answer; A floating-point number is expressed as a basic number, an exponent, and a number base which is usually ten but may also be 2.

Again, just read this: http://en.wikipedia.org/wiki/FLOPS

Collossus wrote:SO WHAT THIS ALL MEANS IS.....
In layman's terms the best way I can describe the speed of a petaflop to you is; A petaflops is equal to 10 times the speed of all the networked computers in the US combined!!!

Source(s):

http://www.itjungle.com/tlb/tlb061708-story01.html

1. That claim should be backed up by a source, because it sounds tremendously stupid.
2. Its the media getting into a hype because they don't understand that its not a heroic national effort to use faster (and more) components. The increase of computing power over time is ridiculously constand progress, there are charts that very accurately predict wen we will reach what and how much its going to cost.

Skgoa wrote: 2. Its the media getting into a hype because they don't understand that its not a heroic national effort to use faster (and more) components. The increase of computing power over time is ridiculously constand progress, there are charts that very accurately predict wen we will reach what and how much its going to cost.

Specifically, Moore's law. Wikipedia link for thee, Collossus.

Collossus wrote:In layman's terms the best way I can describe the speed of a petaflop to you is; A petaflops is equal to 10 times the speed of all the networked computers in the US combined!!!

This is hopelessly wrong. A single decent gaming PC has about 5 teraFLOPS of single-precision and 1.5 teraFLOPS of double precision compute. PetaFLOP-class compute grids are fairly common in industry. Maybe you were thinking of exaFLOPS?

Thanks for the corrections guys! I should have posted a disclaimer that I just relayed that info from someone circulating it here at work.

Starglider wrote:

Collossus wrote:In layman's terms the best way I can describe the speed of a petaflop to you is; A petaflops is equal to 10 times the speed of all the networked computers in the US combined!!!

This is hopelessly wrong. A single decent gaming PC has about 5 teraFLOPS of single-precision and 1.5 teraFLOPS of double precision compute. PetaFLOP-class compute grids are fairly common in industry. Maybe you were thinking of exaFLOPS?

A tenth of an exaflop would mean that the average speed of a networked computer in the US would be in the sub-gigaflop range. The sum total of US computing capacity is certainly in the exaflop range at this point.

Panzersharkcat wrote:

Skgoa wrote: 2. Its the media getting into a hype because they don't understand that its not a heroic national effort to use faster (and more) components. The increase of computing power over time is ridiculously constand progress, there are charts that very accurately predict wen we will reach what and how much its going to cost.

Specifically, Moore's law. Wikipedia link for thee, Collossus.

The only problem with Moore's Law is that it was created and applied to a specific subset of exponential technological progress, when in fact the exponential trend of technological progress is vastly wider than most people realize. It also has produced erroreus conclusions from many people regarding limitations and future projections of technology.