OTOY Forums

Hi there, i have been using octane for a few years now and i´m starting to get the hang of it
but i have ran into a problem. my current GPU only has 1280 mb vram. and compared to everyone here, it sounds like i got an embarrassing low-end graphics card. but ok, i bought it a few years ago so i know it´s already outdated

But, the problem is that i got a polygon limit around 4 million polygons. at that point my vram is used up. of course i can add instances to help with that but for my particular problem it won´t work. currently i´m working on some huge crowd scenes, which means that i will have lots of "duplicated" geometry and it fills my vram quickly.
My current method to overcome this issue is to split the geometry into multiple renders and then comp it together in AE. tedious but it works.
after writing this i came to realize that it´s more like a question rather than a problem. I have read that if you add another GPU to an already existing rig, the performance is increased linearly, correct ?
But what about the Vram ? does that mean that if i got an 1GB vram card and added another one with 1GB vram, i would get the doubled polygon count ? like 8 million poly.
i have been looking for a card for some time now, but i find it difficult to find the right one to the right price.
My thoughts are on the Geforce GTX 970 with 4 GB Vram.

Oh and before i forget it. what does it mean when they say infinite mesh and polygon size ? i read the post about octane 3.0 but i thought the polygon limit was purely limited by the GPU, or am i wrong ?

hope someone here could help me out here. thanks

When adding another GPU, the rendering speed is increased linearly. However, as the entire scene has to be loaded into the VRAM of each GPU, the memory limit for OctaneRender is equivalent to the GPU with the smallest available VRAM.

If you use both a GTX 970 (4 GB) and your current card (1GB), the limit for OctaneRender will be 1GB.
(Other programs using GPU Computing may further reduce the amount of VRAM available to OctaneRender)

The Out-of-core feature available in the development version can offload textures to RAM, however this comes at an expense of render speed.

As for the query about the infinite mesh and polygon size, OctaneRender 2.0 has a 19 million polygon limit.
This is due to a limitation in CUDA that has limited the size usable for geometric primitives to 2GB.

However in the 3.0 version, this OctaneRender engine limitation will be removed.
(so it will only be limited by the GPU hardware)

thank you so much for such a quick answer
yes i had completely forgotten about the thing that you can´t combine two different cards with different Vram. But i didn´t get the answer about the polygon limit.
if i replace my old 1GB card with the new one. does that still mean that i would get four times the amount of polygons ? or is it impossible to determine ?
i´m just trying to put down some numbers, since i haven´t found any benchmarks about Vram litmits.

cool thanks about the 3.0 info. looking forward to see some new stuff there

cheers.