PDA

View Full Version : OpenRT realtime raytracer non-commercial version coming


lightblitter22
08-07-2005, 10:25 PM
mailing list here

http://www.openrt.de/getting_OpenRT.php

from their Siggraph blurb:


On the exhibition floor:

SGI Inc. is showing a demo the 350 million triangle "Boeing 777" model rendered with OpenRT/inView on one of their Prism machines on the SGI booth 1511.
Mercury Computer Systems is showing several OpenRT-driven demos, including inTrace's inView software on a Mercury cluster, a Maxxon Cinema4D OpenRT plugin, an inventor@OpenRT demo, as well as ray tracing on a Cell processor. Check out the Mercury booth number 1155.

Shaderhacker
08-08-2005, 06:13 PM
Very nice. I can't wait to see RRT become the norm for rendering in the years to come.

-M

lightblitter22
08-08-2005, 06:32 PM
Even better, OpenRT has a related project called "SaarCOR", which aims to develop fast realtime raytracing hardware. Project page here:

http://www.saarcor.de/

They already have their prototype hardware running. I hope this gets to the point where you can just plug a SaarCOR based board into your computer and enjoy much faster renders than possible before.

enygma
08-08-2005, 06:34 PM
The nice thing about it, which makes me pretty happy, is they had a paper running the raytracer on the first day of the conference. The session involved a demo where they had programmed the raytracer into Xilink II Vertex 6000 FPGAs running at 66MHz a piece. They were able to program it to scale up to 4 of these FPGAs. I think they were only running it on 2 at the demo and were able to produce caustics and refractions quite nicely at 15 FPS. I think it was a 640x480 display though.

I took a look at a demo at the Mercury booth where they were running on probably 10 to 15 Opteron processors and getting fairly impressive results, but still not what you would want to have for your final frame renders. I don't think the quality of the scenes they were demoing even compared in terms of features to what they showed on those FPGAs.

EDIT
Even better, OpenRT has a related project called "SaarCOR", which aims to develop fast realtime raytracing hardware. Project page here:

http://www.saarcor.de/

They already have their prototype hardware running. I hope this gets to the point where you can just plug a SaarCOR based board into your computer and enjoy much faster renders than possible before.
That is what I was talking about with the FPGAs.

lightblitter22
08-08-2005, 06:39 PM
This is some info off the Saarcor page

PDF (4,8 MB) (http://graphics.cs.uni-sb.de/%7Ewoop/rpu/RPU_SIGGRAPH05.pdf) BIB-TEX (http://graphics.cs.uni-sb.de/%7Ewoop/rpu/RPU_SIGGRAPH05.bib) (full version only in conference proceedings)


The RPU is a fully programmable ray tracing hardware architecture, with support for programmable material, geometry and lighting. The RPU combines the efficiency of GPUs with the advantages of ray tracing. The instruction set of the RPU is GPU like, which is optimal for shading purposes. In addition the RPU supports fast ray traversal through an k-D tree using a dedicated hardware unit and recursive function calls, usefull for recursive ray tracing. To increase efficiency always 4 rays are handled in a packet and multi-threading allows for high utilization of the hardware units.

A working prototype of this hardware architecture has been developed based on FPGA technology. The ray tracing performance of the FPGA prototype running at 66 MHz is comparable to the OpenRT (http://www.openrt.de/) ray tracing performance of a Pentium 4 clocked at 2.6 GHz, despite the available memory bandwith to our RPU prototype is only about 350 MB/s. These numbers show the efficiency of the design, and one might estimate the performance degrees reachable with todays high end ASIC technology. High end graphics cards from NVIDIA provide 23 times more programmable floating point performance and 100 times more memory bandwidth as our prototype. The prototype can be parallelized to several FPGAs, each holding a copy of the scene. A setup with two FPGAs delivering twice the performance of a single FPGA is running in our lab. Scalability to up to 4 FPGA has been tested.

Screenshots are presented 1024x768 resolution with oversampling turned on for most scenes. Please note that all lights, shadows and reflections are calculated and no lightmaps or environmental maps have been used. Detailed measurements of all scenes can be found in the paper.

The following video is computed in realtime on two FPGAs cards.
Video : 36 MB, MPEG-4, 512 x 384 (http://graphics.cs.uni-sb.de/%7Ewoop/rpu/pics/rpu_video_2xfpga.avi)

Hazdaz
08-08-2005, 11:31 PM
I am checking out some of those links now - and this truely sounds like the "holy grail" of CG design....

I find this quote to be the most interesting:
"High end graphics cards from NVIDIA provide 23 times more programmable floating point performance and 100 times more memory bandwidth as our prototype."

Am I reading that correctly?!? That a high-end NVidia card could theoretically be roughly equivalent to a scene from the video that says 32-processors?!?

Oh yea, the scene with the sunflowers was jaw dropping! :eek:

enygma
08-09-2005, 02:17 AM
Well, it is a prototype. I'm just going off of from what I know about FPGA performance, but I think that maybe in order to get the performance they are getting, they are converting floating point to integer. FPGAs have excellent integer performance and with large integer numbers, you can simulate floating point precision.

In regards to memory bandwidth, it is one of Xilinx lower end FPGAs so I wouldn't expect it to be able to push data around like the GDDR3 controller on an NVIDIA card. You may be able to stream through DDR400 speeds with one of the higher end FPGAs. I'll have to see what our guys are doing. Our applications require heavy memory bandwidth, but we aren't using the Vortex II FPGAs.

CGTalk Moderation
08-09-2005, 02:17 AM
This thread has been automatically closed as it remained inactive for 12 months. If you wish to continue the discussion, please create a new thread in the appropriate forum.