15Bit Offline Upload & Sell: Off
|
msoomro wrote:
Little confused. Does it mean using two cards?
Thats the new Nvidia "solution", at least in the professional space.
They are actually segmenting the market very cleverly:
- The GeForce gaming cards are made very powerful for single precision maths and game-type rendering throughput, but don't include optimisations for OpenGL model rendering (or are intentionally crippled, depending on who you ask), and don't support 10 bit colour.
- The Quadro's give you OpenGL and 10 bit colour, but are heavily downclocked relative to their gaming equivalents and are thus slower for single precision Compute (OpenCL and CUDA) than the gaming cards. They are also 4-5x more expensive for the same core GPU.
- The Tesla cards give you single and double-precision compute in the form of a co-processor (no graphics output). To ensure sales of these, both the Geforce and Quadro lines are seriously crippled for double precision maths, except for the GTX Titan and the "Oh-My-God-I-Could-Buy-A-Kidney-For-That-Money" Quadro K6000.
So they have set up their product range such that anyone wanting "workstation" features like 10 bit colour or OpenGL performance has to pay a hefty premium for it, and anyone wanting 10bit colour, OpenGL AND decent compute capability has to buy two expensive cards - a Quadro and a Tesla.
I am not sure how well a Quadro + Geforce setup would work. I would check the documentation carefully before taking that route, as Nvidia might well have crippled it in the drivers.
|