New user hardware query.

Accelerated Computing CUDA CUDA Programming and Performance
1

Completely new to this, but looking to use GPU acceleration of things like MATLAB, ANSYS etc.

As far as I can see high-end gaming graphics cards like the Titan have very similar specs to say the K40.
So why would I buy a K40 rather than the cheaper top end graphics card?

With the K40, I have to have a separate graphics card of course as it has no video outputs.
But if I get, say a Titan primarily for ‘compute’, does it make sense to have a separate, lower spec card to actually handle the video output? A 950 maybe. If I have understood correctly, it will only use the graphics card to accelerate things when it’s not being used for normal video processing load; so my thought was to offload that to a second card, where a cheap one would be perfectly adequate.

Any advice most welcome.

2

This topic has been covered in many places, extensively.

You may be interested in this:

[url]http://www.nvidia.com/object/why-choose-tesla.html[/url]

There are certainly many people who use GeForce products for computing tasks.
You should certainly be able to use a GTX Titan X and GTX 950 in the same box, assuming your platform will support them (slots, power, etc.)