I'm not sure if this is a well known fact, as I just learned this moments ago. I don't have the direct source, so I'm just going to quote the information and provide a link to where the info was given.
full post can be read here: http://www.shacknews.com/chatty?id=26824104#item_26824104 When the OP replies to where he got the information, I will update this post.
UPDATE: The OP says he works for Nvidia and the statement about transcoding is his own words. He said he did not specifically work on the code (he works in developer relations department). He has a long post history and is a trusted member of that community so I believe everything he says.
Transcode basically converts from the disk format to an in-memory format usable by the GPU. GPU Transcode performs this costly operation on the GPU instead of the CPU, which massively increases the number of pages that can be transcoded at a time (while maintaining 60 or near 60 hz). This has the visual effect of reducing the time between an item coming on screen and that item having the full, correct texture data available.
Why is this NVIDIA only?
The underlying code is written in C for CUDA. When the engineer working on this started, OpenCL wasn't an option. Later, preliminary tests into porting the code to OpenCL showed poor performance compared to the C for CUDA implementation.
full post can be read here: http://www.shacknews.com/chatty?id=26824104#item_26824104 When the OP replies to where he got the information, I will update this post.
UPDATE: The OP says he works for Nvidia and the statement about transcoding is his own words. He said he did not specifically work on the code (he works in developer relations department). He has a long post history and is a trusted member of that community so I believe everything he says.
Last edited: