CUDA on multiple GPU

[falcon76]

I would like to understand better how 3DCoat can take advantages of more than one CUDA card on the PC. I have actually two GPU (one quadro FX580 for display and one GTX470) one of them not connected to the display and used only for GPGPU computation (octane rendering in my case).

I'm using only CUDA capabilty of the FX580 actually?

Thanks

Luca

[AbnRanger]

I would like to understand better how 3DCoat can take advantages of more than one CUDA card on the PC. I have actually two GPU (one quadro FX580 for display and one GTX470) one of them not connected to the display and used only for GPGPU computation (octane rendering in my case).

I'm using only CUDA capabilty of the FX580 actually?

Thanks

Luca

I'm not totally positive, but I seem to recall Andrew mentioning early on in the v3 Beta period that 3DC only utilizes one card. That's pretty much the case everywhere in CG apps. with only a few exceptions. It would be nice and maybe Andrew could address that at some point, but I think he feels there are currently bigger fish to fry.

[GeeJay]

I agree, would be nice to have some kind of info about what gets used or maybe even the ability to assign one card to CUDA and the other one for display purposes, if that could enhance performance. But i also think this is rather low priority right now.

[Takkik]

I would like to know too how 3d coat handle multiple cg. I'm projecting to get a second nvidia card for use 3 monitors and so i would like to know how it will be used by the program. is it using the one displaying the program? the one specified as primary diplay or the card in the 'first' pcie slot?

CUDA isnt written for use multiple tesla cards normaly? (perhaps the cg ones can't handle it).

[AbnRanger]

I would like to know too how 3d coat handle multiple cg. I'm projecting to get a second nvidia card for use 3 monitors and so i would like to know how it will be used by the program. is it using the one displaying the program? the one specified as primary diplay or the card in the 'first' pcie slot?

CUDA isnt written for use multiple tesla cards normaly? (perhaps the cg ones can't handle it).

I'm sure it uses the card that is displaying the application. You could still have a 2nd card running a 2nd monitor, but I am not aware that 3DC is yet capable of utilizing multiple CUDA cards.

I think we can easily confuse dual card use for renderers and realtime modeling, sculpting work. They work differently, so it might not be nearly as efficient as you would think. Renderers can simply assign buckets or frames, but for sculpting...communicating between 2 cards might actually be slower.

[falcon76]

I think we can easily confuse dual card use for renderers and realtime modeling, sculpting work. They work differently, so it might not be nearly as efficient as you would think. Renderers can simply assign buckets or frames, but for sculpting...communicating between 2 cards might actually be slower.

I'm not sure about that.

Actually the 2nd card is not used (I check the load with nvidia inspector). I try to force the use with CUDA in the Nvidia control panel, but did'nt work.

But we'll see in the future what's happen.

[Takkik]

Thx guys, so I need to plane for have the best card on the monitor I use 3d coat (important in the case of a possible futur cintiq).

But how does 3d-coat handle a sli config (if it's compatible)? The two cards work on cuda or only one?