DirectCompute anyone?

[Skoshi Tiger]

Well it's like this. When SOW comes out I'll probably get a more modern graphics card to freshen up my PC so I will need a reason to keep my old 9800GTX in my Box and doing something useful.

Oleg has already told us that they are working on their own physics engine (or so I remember) so PhysX support on my old card is out of the question.

I was wondering if there was any scope for DirectCompute support within the SOW series??????

http://www.nvidia.com/object/directcompute.html

cheers!

[Flanker35M]

S!

I really really hope they won't take the road of giving support to only one brand this time, be it red or green team. DirectX 11 capable cards are out from BOTH brands when SoW is released. Please..

[AndyJWest]

Unlikely, I'd have thought.

Firstly, producing specialist support for something like this is hardly good use of limited development time. And secondly, the GPU will be running hard on the graphics anyway!

[Skoshi Tiger]

Being a Microsoft API, I assumed that ATI would would be able to implement it's own version of it at some time otherwise people would just program in plain Cuda. (Could be wrong or ATI may not be interested in it)

Even if it's only used for things like AI such as path finding for ground units where the parallel processing would come in handy and take some cycles off the CPU and allow an environment rich in moving targets it would be a benifit.

Some thing like DirectCompute (like DirectX) would allow the GPU to be used if the spare card is present otherwise the 'standard' CPU intensive routines would be used. (Just like in DirectX)


Not trying to teach Oleg how to suck eggs in any way shape or form, just wondering if we'ld be able to take advantage of all this neat hardware that's been available for the last few years.

Cheers!

[Skoshi Tiger]

Originally Posted by AndyJWest

Unlikely, I'd have thought.

I tend to agree with you!

Originally Posted by AndyJWest

And secondly, the GPU will be running hard on the graphics anyway!

You can have a second graphics card dedicated to run the procedures. from memory anything from 8800 up will run Cuda. And you don't have to mess around with SLI (which isn't really useful at the low res I can show on my 22'' monitor.
Cheers!