NVIDIA Launches Maxed-out GP102 Based Quadro P6000

[AllBodies]

Will either of those be cost-feasible, though? Titan X is already clearly using the chips that weren't good enough for the fully enabled Quadro. New process - probably fairly lousy yields on a big chip this early. Can Nvidia afford to release a cheaper card that uses binned chips?

Yeah they should be. Because they'll be coming March next year at the earliest (going off previous examples).

So by then the 16nm process should be far more mature.

 

[humbug]

Still no HBM2?

 

[Dicehunter]

I wonder what NVidia will call the next fully enabled Pascal GP102 Titan.

I'd guess Titan X Black Edition possibly, It's the reason I'm not getting this Titan XP, No fully enabled chip no buy from me

 

[AllBodies]

Still no HBM2?

Looks like they're skipping it till Volta.

 

[NirK]

Hm so the Titan X is already a GPU with some cuda cores disabled... this means possibly that there will not be a 1080ti.... but they may release one anyway I don't know.

 

[JanM]

Is double precision not needed on quadro cards anymore? What are they using the GP100's that are not good enough for tesla on?

 

[D.P.]

Is double precision not needed on quadro cards anymore? What are they using the GP100's that are not good enough for tesla on?

No, quadro cards are for rendering, CAD work and the like. You are mainly paying for professionally certified drivers and a lots of VRAM.

The GP100 are used for HPC compute and deep learning applications.


Quadro and Tesla are 2 distinct lines of products.

 

[JanM]

Ah ok thanks for clearing that up.