Nvidia, Pegasys announce CUDA support for TMPGEnc

Author: Tim Smalley
Published: 25th August 2008
Comments (12) Stumble
Nvidia and Pegasys have announced that TMPGEnc 4.0 Xpress will support CUDA and deliver some pretty big performance gains.

Nvidia and Pegasys have announced that TMPGEnc 4.0 Xpress will support CUDA and deliver some pretty big performance gains.

Pegasys has announced that it has managed to speed its TMPGEnc video encoder up by as much as 446 percent thanks to Nvidia’s CUDA platform.

The company is demoing a beta version of TMPGEnc 4.0 Xpress, which enables GPU acceleration in the application. The speedups were shown using a GeForce GTX 260 graphics card and a Core 2 Quad Q9450 which by all intents and purposes both fit into the performance segment.

TMPGEnc features support for many popular media formats, including MPEG, AVI, WMV, DivX, FLV and also DVD.

Tak Ebine, CEO of Pegasys said that CUDA has helped to “dramatically improve the filtering speed” in TMPGEnc 4.0 Xpress and then added that because CUDA is based on C, it was relatively easy to achieve these speedups.

This follows on from Elemental’s BadaBoom video transcoding application and shows that it wasn’t just a one-off. One problem that we’ve experienced with BadaBoom so far though is that the video quality isn’t quite as high as we’ve seen from conventional CPU encoders.

With that in mind, we’re going to hang tight until we’ve been able to do a close assessment of Pegasys’ encode quality before making any conclusions about TMPGEnc 4.0 Xpress’ viability. The good thing is that we should be able to get our hands on a final or near-final version of the software soon.

Excited by the prospects of massive video encoding speedups? Discuss in the forums.
Quote Bindibadgi 25th August 2008, 22:39
FOUR HUNDRED AND FOURTY SIX PERCENT

And Tmpeg was one of the things Intel was poster childing as having SSE4.1 support back last November.
Quote bubsterboo 25th August 2008, 23:53
YAY :D I hope to see more CUDA support in more apps!
Quote Icy EyeG 26th August 2008, 00:21
Quote:
Excited by the prospects of massive video encoding speedups?

I am. :D Specially if CUDA would allow me to NOT having to wait overnight for an Adobe Premiere project to encode to a Video file...

I can't wait to see how Adobe is going to implement CUDA on their programs (I do hope they do it on Premiere Elements).
Quote Ramble 26th August 2008, 00:35
Come on AMD. I brought this card because it's bloody awesome, now I want some more uses for it.
Quote Jojii 26th August 2008, 00:40
um space heater?
Quote Cinnander 26th August 2008, 01:06
Quote:
Originally Posted by Jojii
um space heater?

ahahah! You, sir, win the internet. For today, at least
Quote WildTangent 26th August 2008, 01:42
AMD may be ahead as far as hardware goes for now, but stuff like this is what's going to keep Nvidia on top, in my mind anyway. AMD is lagging far behind in the GPGPU segment at the moment.
Quote Phil Rhodes 26th August 2008, 03:13
Careful. The entire purpose of TMPGenc is to be really really damned good.

I can make it 446 percent faster by making it really, really bad.

Need some comparative SNR figures for what must be an entirely new encoder.
Quote mclean007 26th August 2008, 11:05
Quote:
Originally Posted by Phil Rhodes
Need some comparative SNR figures for what must be an entirely new encoder.
Not necessarily - it could be the same encoding algorithm using new code that is optimised for the GPU, so the result would be mathematically identical and hence indistinguishable from CPU encoded video.
Quote kenco_uk 26th August 2008, 12:07
Just imagine, if CUDA was based on assembly language rather than C, the speed advantage would be 1337%.
Quote KayDat 26th August 2008, 14:11
Quote:
Originally Posted by kenco_uk
Just imagine, if CUDA was based on assembly language rather than C, the speed advantage would be 1337%.

Or perhaps OVER 9000!!!


Or not.
Quote wuyanxu 26th August 2008, 14:37
the BadaBoom test on Anandtech didn't show very large improvement in encoding speed from q6600 to 8800GT (or was it 9800GTX, i remember it's a G92 chip)
so IMHO it'd be best to leave Geforce to Fold while use CPU to encode.
Log in

You are not logged in, please login with your forum account below. If you don't already have an account please register to start contributing.





Stats: 0.033 seconds