why cant I use all of my vram on both GPU's?

Sat Jul 25, 2015 1:58 am

I have a 780 ti and gtx 560. if I use both of them to render, for some reason it selects to use the vram of the 560 which is 2GB. if I want to use the vram (3GB) for 780 ti I have to deselect the gtx 560 in the settings menu and then I cant take advantage of its cuda cores.... How do I take advantage of both cards and use all of my vram which would total 5GB and use both graphic cards to render out my scenes? my scene is getting to big and I cant figure out how to use all my vram. Thanks in advance.

Sat Jul 25, 2015 11:08 am

Octane run codes on gpus synchronously over same copy of data. It's parallel computation not serial. Answer is here. Max ram is as lowest ram.

Sat Jul 25, 2015 11:08 pm

aoktar wrote:Octane run codes on gpus synchronously over same copy of data. It's parallel computation not serial. Answer is here. Max ram is as lowest ram.

Ok so if I buy a gtx 970 to go with my 780 ti...Octane for some reason will only use the 3gb of my 780...or will it use 4gb from gtx 970? is it possible this will this be fixed in the future? or is it just the nature of the program? I dont want to get rid of my 780ti but I desperately need bigger scenes..

Sat Jul 25, 2015 11:32 pm

I can't say that "fixable". Because it's not a bug or mistake. It's nature of design to get faster performance. But I suppose that in 3.0 period will be some improvements for vram usage and polygon limits.

Sat Jul 25, 2015 11:45 pm

aoktar wrote:I can't say that "fixable". Because it's not a bug or mistake. It's nature of design to get faster performance. But I suppose that in 3.0 period will be some improvements for vram usage and polygon limits.

Ok, thank you for the info Aoktar.

Mon Aug 03, 2015 7:57 pm

Your computer will always use the lower amount of VRAM because it parallel computes the VRAM from the PCI-e lanes. Stacking is not possible as of yet and even SLI'd it progressively renders every other frame with games etc......The only way I have found to supersede this currently is to make sure you only have the cards with X amount of VRAM needed. I use 2 GTX 980 ti's connected via an SLI bridge, but SLI disabled, it seemingly gets better results via Octane bench 375 compared to 355. When I put one of my extra 960's in with them it automatically reverted to the 2 GB of VRAM but did push my Octane bench to 425........So there are benefits to having more cards, but there are definite drawbacks to not using the same amount of VRAM. Now with DirectX12, theoretically VRAM stacking is possible, but it has to be implemented by the publishers of the software and firmware. Hope this gives a little more understanding.