Build my own workstation?

Accelerated Computing CUDA CUDA Setup and Installation
1

Hello,

I’m looking for information about workstation’s configuration with a double precision GPU (typically the Tesla K40) and i’ve got several questions on this subject:

On Linux, how to manage the fact that i’ve got a GPU to display the graphics and a GPU to compute?
How tell to CUDA which one it have to use?
Which electrical power (yield,power)?
What’s the difference between the product range Quadro and Tesla?

Is there others specificities about motherboard, CPU, RAM?

It seems to me that if we want double precision, we are forced to have a Tesla, no? Because their ratio is the smallest, for example of a GTX Titan X the ratio is 1/32 …

Thanks

2

up?

3

On Linux, how to manage the fact that i’ve got a GPU to display the graphics and a GPU to compute?

You can configure your X server. But I think it will automatically configure for the first accelerator with
actual video modes available.

How tell to CUDA which one it has to use?

using the variable: CUDA_VISIBLE_DEVICES

Which electrical power (yield,power)?

I think you have to consider 2000W

Do not take those answer as an absolute truth!