Page 2 of 2
Re: OpenRL
Posted: Tue Mar 23, 2010 8:54 pm
by radiance
i don't think i can use this library, octane runs completely on the GPU.
this library would be for accelerating CPU based engines by offloading only the ray intersections to the gpu,
which has already been proven to marginally increase speed. (2x-3x)
Radiance
Re: OpenRL
Posted: Tue Mar 23, 2010 9:13 pm
by [gk]
Well you could adabt Octane to gain access to any kind of hardware to drive it via openRL.
I know brazil was ported to orl and working inside 3dsmax rt, ofcourse as you mention its an offline render from start point, but I wouldnt think its impossible to port Octane even though its optimized for the gpu alone, excluding the cpus in the system in addition any other hardware.
Maybe take a talk with caustic guys, they know the answers. Sometimes better than guessing.
Its worth it I think thinking long span here, not short term goals in mind.
Also worth to note that a possible fermi when the issues has been solved consumes a ton! of power over the caustic ray card 30w. ( this is again another aspect as there is both the raycard accelerator and then the new API )
Cards as the fermi are hybrid cards with some new additions to the architecture where the caustic2 card is a pure raytrace card.
Ofcourse a ray test of the 2 types would have to be made in junction with their costs and not to forget how much power they needin order to determin the profitable pick for the clients..
And if someone makes some new piece of hardware it can slide directly into the render bridge and support.
Re: OpenRL
Posted: Tue Mar 23, 2010 9:51 pm
by gpu-renderer
that caustictwo ray trace card seems a better idea than just gtx cards... the wattage, single slot design will certainly works wonders... The ram slots on them 2 isa good idea. Lots of cheap ram for the job. I came across a evga board with 7 single slot pci gen 2 slots and dual xeon with upto 48gb of ram. 6 x caustictwo's + main gpu and dual xeons would be one serious work horse. Also the specialized aproach would be faster and more accurate.. time will tell though and the cost factor of these boards... (hate to think of the price). But running costs are good looking at 800 watts for all of that. 120x speed of gtx 285 if your willing to pay the tab :0)
Anyone know any more details.
Re: OpenRL
Posted: Tue Mar 23, 2010 10:17 pm
by [gk]
Well you probably already surfed their site
But here are some other videos they made. ( for those who would like to take a peek )
http://caustic.com/gallery_movies_globa ... nation.php
Re: OpenRL
Posted: Thu Apr 01, 2010 12:55 pm
by [gk]
Re: OpenRL
Posted: Fri Apr 02, 2010 3:30 pm
by pedrojafet
radiance wrote:i don't think i can use this library, octane runs completely on the GPU.
this library would be for accelerating CPU based engines by offloading only the ray intersections to the gpu,
which has already been proven to marginally increase speed. (2x-3x)
Radiance
"...completely on the GPU."
My CPU process are used a lot while rendering in Octane...
Is it normal, or my GPU driver aren't doing the right job?