Thanks a lot Mik for the advice :)

What a headache !!!

I can't do it anymore :)

I've read so many tech articles in the last two weeks...

Cuda, GFlops, Fp16,Fp32,Fp64, TFlops, ECC Ram vs Non-Ecc, Quadro Kepler, Quadro Maxwell, Titan Black, Pascal....

So many!

It'a a jungle :)

After read all that stuff I've come to realize something :

1) NVIDIA after Fermi cards has decided to render ineffective all FP64 (double precision) performance of ALL it's cards,
both Quadro and GeForce!
In fact even the powerful Pascal Quadro P6000 equipped with the GP104 processor line and the Gtx 1080 Ti equipped with the GP102 (same family)
have FP64 performance fixed to 1/32 of FP32.

2) The "old" Gtx Titan Black from 2104 is equipped with the GK110-430-B1 and has a FP64 performance varying from 1/24 to 1/3 of FP32 !!!

But...whit that said....I didn't find any place or person who can give me this simple information :

Considering that I don't need the GPU to process massive scientific calculation (FP64) but only for visualization, 3d modelling, 3d rendering and
2d Gpu Painting, can I get rid of FP64 performance and go for a ultra powerful FP32 (11 TFplos) Gtx Nvidia 1080 Ti ?

No one has still told me...

I think I should go for a Gtx 1080 Ti and get Rid of Pascal Quadro because they use the same chip with the almost same FP64 performance.

Only, it seams that without a Quadro card I will loose 10-bit display output and ECC memory.

As I said...it's a Jungle!

Have a nice day to everyone.

- Marco (mkdm)