CUDA encoder has been updated and optimized. Compiled against CUDA 5.0, frames pre-processed in NV12 colorspace, process loop refined, better GPU utilization and better performance can be achieved with CUDA encoding. More options are available on MediaCoder’s CUDA tab, including the one specifying the GPU device to be used for encoding when there are multiple ones.
Benchmarking shows that the GPU utilization can reach 82% on a GTX480 whhen Offload option (newly added) set to Full. The speed of a typical 720p transcoding reaches 8.6x.
So, stop watching, grab the update, try on your GeForce, see the difference and let me know if there is!
Hello. I installed this update. Cuda encode works fine with previous version. But i receive error code 14 with this version now. I think my settings are true. My graphic card is gtx560ti. Last drivers installed. And i use Windows 8 x64. I want to use this cuda options. Please help me. Thanks.
Make sure to use 64-bit MediaCoder on 64-bit Windows. Otherwise, CUDA encoding may fail to work. Also I noticed baseline profile sometimes does not work, so you can try main or high profiles.
Thank you for your answer. I installed 64bit MediaCoder. I’m sure about that. I tried different settings. Still not working.
I too have a GTX560TI card and am running Win 8 (x64). I get error code 14 as well with NVIDIA 310.70 (x64) drivers and 64 bit MediaCoder (0.8.18.5338) installed. I previously had MediaCoder-x64-0.8.16.5295 installed with NVIDIA 302.82 (x64) drivers and all was working well.
Either the CUDA v5.0 support in the NVIDIA 310.70 drivers is not working or there’s a bug in MediaCoder v0.8.18.5338 (x64).
An update will be released soon for this issue.
RU У меня вопрос к разработчику, можно как то улучшить качество при кодировании CUDA, сделать в два прохода? GTX780 @1.2GhZ справляется быстро, но качество по сравнению с кодированием без CUDA в 2 прохода, хуже.
EN My question to the developer, as it can improve the quality of the encoding CUDA, done in two passes? GTX780@1.2GhZ cope quickly, but the quality compared to coding without CUDA 2-pass, worse.