PDA

View Full Version : Nvidia Cg download location?


teatimebing
07-30-2012, 10:11 PM
Seems like Nvidia are still restoring the developer zone. Does anyone know where I can download the latest Cg toolkit for Mac and Linux?

Is it included with another download?

This is all I could find:
http://www.nvidia.com/content/devzone/index.html

djwarder
07-31-2012, 04:09 PM
If I'm assuming v1.5 is the latest, this should work ...

http://www.nvidia.com/object/cg-toolkit-15#downloads

teatimebing
07-31-2012, 04:28 PM
It's not the latest, but you did inspire me to change my google syntax and I found this page!:

http://http.developer.nvidia.com/Cg/index_releases.html

And this is the latest:
http://http.developer.nvidia.com/Cg/cg_3_1_0013.html


EDIT:
Updated, this seems to be the current URL:
http://developer.nvidia.com/cg-toolkit

djwarder
07-31-2012, 05:01 PM
Aha, wondered if it wasn't the latest. Not used Cg, thought about but then moved to CUDA/OpenCL.

teatimebing
07-31-2012, 07:24 PM
Interesting! I had been looking at OpenCL and CUDA, I have been trying to make some kind of wrapper for crunching queries similar to what I can achieve in SQL now. But of course I would have to get the whole dataset into VRAM first, then chunkify it for distribution of the trillions of cores your average GPU has for stream processing similar to MapReduce methods.

My current interest in Cg is for shader development, but after going through the documentation I am immediately seeing the benefits of crunching on the GPU in what I think to be a very accessible platform.

This was an interesting read http://www0.cs.ucl.ac.uk/staff/ucacbbl/cigpu2008/slides/shader-vs-cuda.pdf

djwarder
08-01-2012, 11:32 PM
Intersting article. Still getting my head round OpenCL at the mo, but is sooo powerful if you get it right!

mje11even
08-13-2012, 09:06 PM
Just a word of caution right now about using CUDA/OpenCL for things other than graphics display. I have found that the cost of uploading and downloading data to and from the GPU can wipe out any performance gains from all those cores doing the work on the GPU. Your calculations need to be computationally intensive vs. the amount of data input and output. It's worth doing some quick profiling in CUDA to check the time it takes to copy your data to and from the GPU without doing any calculations vs. just doing the calculations on the CPU.

teatimebing
08-16-2012, 10:16 AM
Where are you staging your data though? In RAM or on spinning disks?

An advantage of the Mercury/Fixstars CELL PCIE boards was dual-gigabit ethernet straight into the device.

CGTalk Moderation
08-16-2012, 10:16 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.