OTOY Forums

I didnt know where this belonged so I just stuck it in here. I figured it would be more helpful to ask an experienced community.

Well, Im pretty new to a lot of this unbiased 3D stuff. But I was drawn to octane because of its quick ability to render realism. I had a few important questions and was hoping someone could shine some light on this...

First, my reason for choosing octane is because Im going to be making my own indie films using live action camera's and I want to incorporate rendered 3D into my scenes for visual effects. Im a blender user and i tried blender internal as well as yafaray but they dont give me the level of realism Im after. Not only that but I dont feel like spending a ton of money on pc's to set up a render farm. Thats why in my case something like octane would be a godsend.

So heres what Id like to know (forgive me for asking so many questions but just incase these topics havent been touched on yet I wanted to get all the info I can in one shot. On a sidenote Im not a complete noob but I am a little uneducated in some area's)...

1.) Can octane be used for motion graphics/visual effects? I mean after the scene has been tracked the camera movement is already established right? So all I would have to do it export that tracked data animation to octane? Im assuming.

2.) Whats the difference between GPU cores and CUDA cores? I have researched this but didnt get the answers I was looking for. I read that octane is engineered to work with cuda, and my pc has 2 GTX 260 cards along with core I7 920 and 3 of those heavy duty memory sticks you plug into the motherboard (I dont know if those do anything for graphics, just saying).

3.) If these cuda cores are the magic potion to getting the most power out of octane is it possible to set up a group of maybe 1 or 2 other towers, each one equipped with like 2 cuda enabled cards and then link those to my main PC with a KVM switch so it will act as a render farm so to speak? I mean that wont require any programming will it? User friendly would be very nice.

4.) I was sold on Tesla C1060 until I saw the price. It has 240 streaming processor cores wheres the GTX 260's I have contain 192 cuda cores. So thats why Im curious as to how much more effective streaming cores would be as opposed to cuda cores.
Will the tesla do more for an octane render than a fully equipped cuda card? It doesnt look to me like the tesla contains any cuda cores.

5.) Last, Im kind of embarassed to mention this being a 3D artist and all but Ive noticed that octane has to be configured with nodes before it will render. Well, Im not good with nodes. Matter of fact thats one of my weakest area's. So are there any provided presets I can use to get photorealism for my movies? Like indoor and outdoor node set ups?

Just to add to that can anyone tell me what would give me the power to render path-tracing in seconds? I know thats an unrealistic way to think I just wanted to ask what would be the best way to go about something like that.

I think that about does it. Im open to all comments positive and negative. Either way I appreciate whoever reads this and thanks a bunch for any insight.

Yes Octane works well with tracked footage and you can easily do various passes to generate masks, shadows and so forth.
Howeever as it seems you are semi rookie, expect a fair deal of hard training before you reach to the point where all this feels logic and natural.

You want to get a card with a decent amount of cudacores, it sounds like you think there are different things in a gfx card, cuda cores and something else, there is only cores. Get as many as you can afford and depending on the cg complexity memory is important. Usually way more important that the number of cores available.

Currecntly you cant render over network. So you are limited to whatever amount of cards you can stick in a workstation or the pcie expanders.

Dont go tesla unless you got the money. get one of the 400 range gtx cards.
Best one is the 460 with 2gb memory over all.

Regarding the nodes, well. Work hard and learn to understand the concept, it has been long delayed in the industry and I sure hope every app will be node based and remove the silly old ui clutters.
Nodes are much more efficient to work with, fast to share groups of sub shading mechanisms with a simple drag and drop and so forth.

Hope this helps

[gk] wrote:Yes Octane works well with tracked footage and you can easily do various passes to generate masks, shadows and so forth.
Howeever as it seems you are semi rookie, expect a fair deal of hard training before you reach to the point where all this feels logic and natural.

You want to get a card with a decent amount of cudacores, it sounds like you think there are different things in a gfx card, cuda cores and something else, there is only cores. Get as many as you can afford and depending on the cg complexity memory is important. Usually way more important that the number of cores available.

Currecntly you cant render over network. So you are limited to whatever amount of cards you can stick in a workstation or the pcie expanders.

Dont go tesla unless you got the money. get one of the 400 range gtx cards.
Best one is the 460 with 2gb memory over all.

Regarding the nodes, well. Work hard and learn to understand the concept, it has been long delayed in the industry and I sure hope every app will be node based and remove the silly old ui clutters.
Nodes are much more efficient to work with, fast to share groups of sub shading mechanisms with a simple drag and drop and so forth.

Hope this helps

That helps me a lot. Thanks that pretty much sums up what I wanted to know.

one more questions to ask... Ive been reading around that the poly performance in real time is determined by the amount of vram available. Does this mean that the only vram thats going to be used is the amount on the card that the monitor plugged up to, or is all the vram from every card in the motherboard utilized?

thanks again gk, that was a big help.

Only memory solded to your gfxcard(s) will be available through Octane.
So no system memory.
If you have a gtx260 with about 900mb memory on and a gtx480 with about 1.5gb memory on, the available memory pool in octane will be the gtx260s 900mb.
Or simpler put, The weakes link is your workspace. This si why it is important to get 2 or more of the exact same card when you purchase.

Do note that you can use a lowend card to drive the UI of octane and a fine card to render with. This will also clear up performance speedwise working within the UI.

[gk] wrote:Only memory solded to your gfxcard(s) will be available through Octane.
So no system memory.
If you have a gtx260 with about 900mb memory on and a gtx480 with about 1.5gb memory on, the available memory pool in octane will be the gtx260s 900mb.
Or simpler put, The weakes link is your workspace. This si why it is important to get 2 or more of the exact same card when you purchase.

Do note that you can use a lowend card to drive the UI of octane and a fine card to render with. This will also clear up performance speedwise working within the UI.

does that mean the combined memory of that card if theres more than one? for example gtx 260 has 900, so 2 x 900 would be 1800MB from both cards

No you still have 900 total to use. Octane loads the same scene into both cards.