Tesla C2075

[Leme3D]

Somebody already used a card with these octane? Suggestions are welcome ...

[pixelrush]

Unless you already own one don't bother.
You would need to use a Quadro as a display card for the Tesla too.
You can buy much cheaper and also faster Geforce cards.
GTX580/3gb are popular for Octane. 3gb is a decent space for more demanding scenes like for archviz.
A new Octane beta is due soon that will support instancing and allow you to pack more polys into a smaller space. Trees and shrubs, cars in parking lots, or stadium/auditorium seating come to mind.
The Nvidia Kepler generation are due in a few months - the double the vram versions should follow soon after.
Personally I would wait for those and get two GTX680/4gb and say a basic GTX640 for display, all for about $250 less than one Tesla.
This combo will be at least 5x faster at rendering than with the C2075.
The only reason to buy Tesla for Octane is if you must have 6gb vram.
Current hardware limits the number of textures available for Octane to the same number regardless of vram.
To fully feed 6gb with Octane data though you will need a mobo with 32gb sys ram.
For 4gb vram you will need 24gb sys ram.
HTH

[Leme3D]

Thanks for helping me.
My intention is to create virtual tour for architecture.
I have not done a test with plates and even scenes, but I can say that for my projects, I need something sturdy because of the large number of objects in the scene.
For example: The physical body of the building, furniture, cars, plants, grasses, complete leisure area (pool, playground, soccer field, sports court, walking track, barbecue) in purpose, a tour where I'll have to show all these and sometimes even a condominium homes.
Because it is something a little abused over polygons and props I thought these cards with 6GB ...

I got a close partnership with a builder that every four months is launching a new venture.

Now if someone has experience with something like that and can tell me where the memory usage of the cards in scenes so let me know it may be that noa is necessary to purchase such cards.

About the boards mentioned that our friend pixelrush (gtx 680) already have some confirmation on the amount of memory that they will provide? 4GB would be even as he informed me??

I'm a big bang marshal mobo, bought especially for providing me with 8 pci and this could put eight plates to have a faster rendering.
Another thing I would reply, is about the octane of the PMC.
By the time I could not see any kind of tour using this type of tool.
If anyone has any links I pass.
From what I understand the images and some information I got was that the PMC is the most suitable for this type of work.
If I'm wrong please let me know ...

The render time using this PMC is so fast as to what I see in the statements of octane or it is not so fast like that?

Thanks to all who pass by here and leave your msg.

[t_3]

I'm a big bang marshal mobo, bought especially for providing me with 8 pci and this could put eight plates to have a faster rendering.

by the way, in the latest octane manual refractive notes, that a pcie x1 slot isn't sufficient for octane - don't know if anybody already had real-world experiences with this, but as four of the x16 slots of the big bang marshal support in fact only x1 speeds, this might be a problem...

[pixelrush]

I think you might be overly ambitious to be touring with a level of detail of grass and such although instancing will certainly help.
Remember Octane only allows a limited number of textures too but of course you can reuse them any number of times. My guess is that 6gb might hold 45m polys and textures *but* you have to get them in there.
That is you have to set up a scene in an application like Blender or Max and then export it, and then load it into Octane. At a guess you might even need 96gb of sys ram to load that into Octane. I am not sure about a mobo for that - probably need a server board that might have only one pcie 16 slot....
In practice I think you will find it difficult to work with anything over say 15m poly not including instances regardless of your hardware. A 4gb card might be more appropriate in that case.
Probably for a walk around tour you would use directlighting which is simpler, faster but not quite the same image quality. PMC is slow but for some situations it is useful as the fireflies clear quicker. Really I think you would want something faster than a single C2075 for rendering if you have a client watching and you are navigating around a scene with a spacemouse.
There is some new technology - out of core memory paging - that allows loading of scene data on demand ie the data for the present view only. At some stage Octane may get this as an option and you could work with many many polys as you propose. See here for an example http://www.centileo.com/news.html This would be a better solution for you but I think the materials and lighting are a bit limited ATM. There again though the card doesn't have so many gb vram but speed would be more useful. For the centileo raid0 ssd or pcie ssd might be handy as well.
HTH.