Double precision

[Nexo]

What is Double precision I really don't know. I goggled it and I couldn't find anything about it.

[wng_kingsley7]

In what context or subject topic ? @Nexo

Maybe tracking for a mousse to be more accurate. You need to elaborate more on what you're trying to say and comprehend unknown meanings.

[FlamingDefibs]

In what context or subject topic ? @Nexo

Maybe tracking for a mousse to be more accurate. You need to elaborate more on what you're trying to say and comprehend unknown meanings.

 

I'm guessing that since this is in Graphics Cards, he may be talking about the GTX Titan.

[Nexo]

I'm guessing that since this is in Graphics Cards, he may be talking about the GTX Titan.

I'm talking about the GTX Titan. I keep hearing things about double precision on it but not knowing what it is. Like what is double precision on a GPU?

[wng_kingsley7]

I'm guessing that since this is in Graphics Cards, he may be talking about the GTX Titan.

Ahh.. So okay in that context the double precision is referring to the larger 6Gb frame buffer that only the geforce gtx titan for running larger a monitor or a surround setup with resolutions over 4K which is 3840*2160

[Mecami]

The 780 has very crippled double precision performance, like the rest of the Kepler GeForce family.  The Titan and Quadro lineup aren't.

 

Double precision does not relate to error correction or accuracy, this is a misconception.  It is about the precision of the floating points.  Double precision floating points are twice as many bits as single precision floating points.  In a single precision world, suppose you had 4 digits, 1.001 would be as precise as you could get.  If the actual value or result of a calculation was 1.001281 then it would still turn out as 1.001 and feed that into the next part of the calculation.  Whereas double precision floating points have twice as many places, and so can be more precise in specifying numbers and holding values.

 

Naturally this results in longer calculations, more VRAM usage, and higher file sizes, and most applications take no benefit from it, so it's not commonly implemented.  CAD and 3D modelling benefit from it most.

 

Original post in here: http://linustechtips.com/main/topic/52299-double-precision-on-gtx-titan/

[Ciccioo]

Ahh.. So okay in that context the double precision is referring to the larger 6Gb frame buffer that only the geforce gtx titan for running larger a monitor or a surround setup with resolutions over 4K which is 3840*2160

...no

 

 

that double precision performance stands for the capability of the titan to work very very fast with numbers with high precision, a lot of decimal digits

double precision numbers are stored in 64bits (http://en.wikipedia.org/wiki/Double-precision_floating-point_format) instead of the usual 32

[SkilledRebuilds]

DFPP is mainly used in the GPGPU applications found in scientific research etc. It is for where calculations that can be made in a parallel fashion also require high precision.

In terms of games, it isn't massively utilized as there is rarely any need to make such precise calculations.

Also FP64 instruction set based calculations massively impact performance. As a general rule, games tend to get away with FP32 (single floating point) instruction set calculations.

[Nexo]

Original post in here: http://linustechtips.com/main/topic/52299-double-precision-on-gtx-titan/

So what is it good for besides 3D modelling? 

[Nineshadow]

Floating points are weird.

Why the heck would you think 0.1 + 0.2 = 0.30000000000000004.

;_;

[Mo5]

Floating points are weird.

Why the heck would you think 0.1 + 0.2 = 0.30000000000000004.

;_;

[Nineshadow]

I watched it.It's still weird.

[Glenwing]

So what is it good for besides 3D modelling? 

 

Not much really.  Scientific research and simulations come to mind.