CUDA performance 980 vs 960 - same?

[idieorudie]

On the GeForce website the compute power of 980 and 960 is given as same - 5.2

980 is wayyyyyyy better but why is it so that the compute performance is the same?

[Mo5]

On the GeForce website the compute power of 980 and 960 is given as same - 5.2

980 is wayyyyyyy better but why is it so that the compute performance is the same?

What is the unit of this magical number - FLOPS?

[Atmos]

I call bullsweet. The 980 is leap years beyond the 960, no way their compute performance is the same. Oh no guys, it's another piece of false advertisement QUICK SOUND THE ALARMS IT'S TIME FOR FLOPGATE!!11!!1!!11

[idieorudie]

What is the unit of this magical number - FLOPS?

I'm confused too!

They just stated 5.2 as compute power!

[Mo5]

I'm confused too!

They just stated 5.2 as compute power!

Can you like the site. I can't find it anywhere

[idieorudie]

Can you like the site. I can't find it anywhere

I'll post screenshot

[idieorudie]

Wtf NVIDIA!!!??

[Manny Calavera]

5.2 is maxwell cuda or mth?

[idieorudie]

5.2 is maxwell cuda or mth?

Mth?

[idieorudie]

@LinusTech can you please explain this?

[Mo5]

sketch-1424967820330.jpg

Wtf NVIDIA!!!??

And Titan Z has the same as Titan which is lower than 960. This makes 0 sense.

[EpIcPoNaGe]

"The compute capability of a device is represented by a version number, also sometimes called its "SM version". This version number identifies the features supported by the GPU hardware and is used by applications at runtime to determine which hardware features and/or instructions are available on the present GPU.

The compute capability version comprises a major and a minor version number (x.y):

Devices with the same major revision number are of the same core architecture. The major revision number is 5 for devices based on the Maxwell architecture, 3 for devices based on the Kepler architecture, 2 for devices based on the Fermi architecture, and 1 for devices based on the Tesla architecture."

Read more at: http://docs.nvidia.com/cuda/cuda-c-programming-guide/index.html#ixzz3SrzRN3FO

Follow us: @GPUComputing on Twitter | NVIDIA on Facebook

[Nikolithebear]

sketch-1424967820330.jpg

Wtf NVIDIA!!!??

WHAT.

[idieorudie]

WHAT.

NVIDIA needs to be clear on their site where they mention these things

[idieorudie]

"The compute capability of a device is represented by a version number, also sometimes called its "SM version". This version number identifies the features supported by the GPU hardware and is used by applications at runtime to determine which hardware features and/or instructions are available on the present GPU.

The compute capability version comprises a major and a minor version number (x.y):

Devices with the same major revision number are of the same core architecture. The major revision number is 5 for devices based on the Maxwell architecture, 3 for devices based on the Kepler architecture, 2 for devices based on the Fermi architecture, and 1 for devices based on the Tesla architecture."

Read more at: http://docs.nvidia.com/cuda/cuda-c-programming-guide/index.html#ixzz3SrzRN3FO

Follow us: @GPUComputing on Twitter | NVIDIA on Facebook

What is the second number?

[EpIcPoNaGe]

NVIDIA needs to be clear on their site where they mention these things

Well to take Nvidia's side, they state "compute compatibility" not compute performance. Also the average use wouldn't need to care about such details... It only took me a few minutes to find that snippet in the above post.

[EpIcPoNaGe]

What is the second number?

Not too sure. Most likely it specifies more specifically what features it has.

[idieorudie]

Well to take Nvidia's side, they state "compute compatibility" not compute performance. Also the average use wouldn't need to care about such details... It only took me a few minutes to find that snippet in the above post.

I am a beginner in 3D work so I got confused but they should have mentioned it there and then!

Can you tell me what the second number means?

[idieorudie]

Not too sure. Most likely it specifies more specifically what features it has.

So it also doesn't denote the power?

Then what does?

[EpIcPoNaGe]

So it also doesn't denote the power?

Then what does?

Here is the chart specifying what the numbers mean.

http://docs.nvidia.com/cuda/cuda-c-programming-guide/index.html#compute-capabilities

[EpIcPoNaGe]

A better representation of compute power would be the flops performance of each card.

Edit:

For example almost all (if not all)the other cards other than the titans have a limited double precision performance. Thats why they market the Titan for more workstation work.

[idieorudie]

A better representation of compute power would be the flops performance of each card.

Edit:

For example almost all (if not all)the other cards other than the titans have a limited double precision performance. Thats why they market the Titan for more workstation work.

Okay thanks!

[EpIcPoNaGe]

Okay thanks!

No problem