OTOY Forums

Hi

I am trying to render some simple stuff on my mac osx with gtx 680 (2GB). in live viewer everything easy. but when i go to final render i cant get renderings with more than 3200 pixels. otherwise i get an error.
its only a couple of mapped cubes. there is more in the scene but switched off.

i know memory is not too high, but can this really be an issue?

best regards
volkmar

if You use Your card to power up the screen, that uses some vRam, - don't know how on MacOS, but in windows that about 300mb for 1080 screen. Then You have other pas possibly using some of that memory. Octane needs to fit models, textures, etc. & image buffer (or however it is called =) basically bigger resolution, more of vRam You need to be able to render. For the reference when I render something like 8K square Octane eats 2GB just for that (scene textures & other expenses excluded =) so..I think it's time to look for better card mate =) 4Gb seams reasonable price, You can get older Kepler or brand new Maxwell (but keep in mind these at least at the moment don't perform the best =) discontinued 6GB 780 was a bargain, try to look for some, otherwise You're left with almost twice as costly TitanBlacks =) hope that explains things a bit =)

addional note, I think that macos consumes lot of vram. don't you check free vram in live viewer? what's the free ram here?

aoktar wrote:addional note, I think that macos consumes lot of vram. don't you check free vram in live viewer? what's the free ram here?

hi aoktar

yes i could check that. i noticed that all open programs consume quite a lot of vram. especially photoshop.
after closing all this i had about 1.2 gb vram free. i could render a full screen live preview (2500 x something) with about 250 mb free. but rendering for a final image is only possible up to 3200. so it seems the resolution adds up quite severely to vram.

so maybe i try to downsize the geometry a bit more. maybe that helps.

cheers
volkmar

You should think to decrease texture sizes rather than geometry. Textures consumes much more memory else geometry.

The best solution would be not to optimise model/textures (as that might be visible in the end), but simply to render the image with lens shift & combine the final result from 2 or 4 shots. Speed wise it will take the ~ the same amount of time, but rendering lower res images (portion of final) will use less of vRam. This way You can get away from limitation of vRam, but still keep relatively high mesh count or detailed textures (remember that on bigger picture more details are going to be visible too).

I suggest the same solution as Glimpse but with the script xpresso 'tiled camera' provided with Cinema 4D in mediatheque.
Don't forget to set vignetting to zero with these techniques.

atome451 wrote:I suggest the same solution as Glimpse but with the script xpresso 'tiled camera' provided with Cinema 4D in mediatheque.
Don't forget to set vignetting to zero with these techniques.

thanks that does the trick!
finally i have a workaround for this problem until i get my workstation.

cheers
volkmar