What is a Teraflop?

[resurrection64]

What is a Teraflop, which graphics cards consist of?
Please explain!

Much appreciated! Thanks!

 

[ingtar33]

it's when Tara Reid parties too hard, passes out and falls on her face.

oh wait, you wanted to know what a teraflop was... not a taraflop.

FLOP - FLoating-point Operations Per-second
so a TeraFLOP = TFLOP = one trillion floating point operations per second.

 

[smeezekitty]

1000000000000 floating point operations per second.

 

[resurrection64]

it's when Tara Reid parties too hard, passes out and falls on her face.

oh wait, you wanted to know what a teraflop was... not a taraflop.

FLOP - FLoating-point Operations Per-second
so a TeraFLOP = TFLOP = one trillion floating point operations per second.

LOL! :lol: What are they for?!

 

[smeezekitty]

Measuring RAW computing power

 

[unksol]

it's when Tara Reid parties too hard, passes out and falls on her face.

oh wait, you wanted to know what a teraflop was... not a taraflop.

FLOP - FLoating-point Operations Per-second
so a TeraFLOP = TFLOP = one trillion floating point operations per second.

LOL! :lol: What are they for?!

It's a unit of measure... So...

 

[resurrection64]

Measuring RAW computing power

So the higher, the faster your GPU can process the video right?

 

[smeezekitty]

In simplified terms, yes.

 

[ingtar33]

it's when Tara Reid parties too hard, passes out and falls on her face.

oh wait, you wanted to know what a teraflop was... not a taraflop.

FLOP - FLoating-point Operations Per-second
so a TeraFLOP = TFLOP = one trillion floating point operations per second.

LOL! :lol: What are they for?!

Understand i'm not a physicist or mathematician... nor am i a computer programmer... so my understanding of FLOP might be wrong... but as i understand it.

Floating Point Math is basically a type of math dealing in approximations... small and accurate approximations, but approximations none the less. It's simplifying a great long (or endless) number (like pi) down to a smaller number with smaller characters, for example, if you were to use 3.14 as a standin for PI in a math problem, you're basically practicing floating point math. The difference between standard calculus (when it's acceptable to limit numbers to a set number of significant digits) and typical computing math, is a computer typically makes no judgment on the number of significant digits in a math problem and calculates to it's programmed limit of numbers.

Floating point is different in that the computer will be making a judgment of sorts as to how many digits in each "number string" are needed for acceptable levels of accuracy, and will adjust the number of significant digits accordingly (as opposed to a set number, the number of significant digits in floating point math... floats on a case by case basis).

The result of this is Floating point is IDEAL for graphics cards and video rendering as absolute accuracy isn't needed, and a great deal of time is spent calculating "triangles" and "curves" or in other words "calculus" over and over again... Hence "TFLOPS" becomes a measure of a GPU's raw number crunching power, and therefor a type of measurement to the general strength of a GPU's core.

This is why gpu cores make such interesting supercomputers... they're uniquely designed to be highly efficient giant calculus calculators

 

[resurrection64]

it's when Tara Reid parties too hard, passes out and falls on her face.

oh wait, you wanted to know what a teraflop was... not a taraflop.

FLOP - FLoating-point Operations Per-second
so a TeraFLOP = TFLOP = one trillion floating point operations per second.

LOL! :lol: What are they for?!

Understand i'm not a physicist or mathematician... nor am i a computer programmer... so my understanding of FLOP might be wrong... but as i understand it.

Floating Point Math is basically a type of math dealing in approximations... small and accurate approximations, but approximations none the less. It's simplifying a great long (or endless) number (like pi) down to a smaller number with smaller characters, for example, if you were to use 3.14 as a standin for PI in a math problem, you're basically practicing floating point math. The difference between standard calculus (when it's acceptable to limit numbers to a set number of significant digits) and typical computing math, is a computer typically makes no judgment on the number of significant digits in a math problem and calculates to it's programmed limit of numbers.

Floating point is different in that the computer will be making a judgment of sorts as to how many digits in each "number string" are needed for acceptable levels of accuracy, and will adjust the number of significant digits accordingly (as opposed to a set number, the number of significant digits in floating point math... floats on a case by case basis).

The result of this is Floating point is IDEAL for graphics cards and video rendering as absolute accuracy isn't needed, and a great deal of time is spent calculating "triangles" and "curves" or in other words "calculus" over and over again... Hence "TFLOPS" becomes a measure of a GPU's raw number crunching power, and therefor a type of measurement to the general strength of a GPU's core.

This is why gpu cores make such interesting supercomputers... they're uniquely designed to be highly efficient giant calculus calculators

Thank you!