NVIDIA CUDA Delivers 446% Speed Increase to Pegasys Video Processing Solution

[btarunr]

Today, at the NVISION 2008 conference, NVIDIA Corporation in conjunction with Pegasys Inc., makers of TMPGEnc 4.0 XPress multi-format video encoding software, showcased a technology demonstration to optimize video processing with the massively parallel architecture of the GPU.

Using NVIDIA CUDA technology (a C-like programming language programming for the GPU), Pegasys is taking advantage of the parallel processing capabilities of an NVIDIA GeForce GPU to create a GPU-enabled beta version of TMPGEnc 4.0 XPress software. The software is used to dramatically increase video decode and processing speed by as much as 446% on a GeForce GPU.



"Leveraging NVIDIA CUDA technology to accelerate our application on the GPU has dramatically improved the filtering speed of the TMPGEnc 4.0 XPress software," said Tak EBINE, CEO, Pegasys Inc. "CUDA technology has helped us deliver this result in a relatively short development time because it is intuitive to C programmers."

TMPGEnc 4.0 XPress software converts and compresses (encodes) all types of video files that can be played on the PC, including MPEG, AVI, WMV, DivX, FLV, as well as DVD video..Pegasys' unique Video Mastering Engine's interface has gained a reputation for being user-friendly, enabling easy editing and conversion of video sources.

"Pegasys' video transcoder software has earned top ratings in Japan and overseas for its quality and ease of use," says Patrick Beaulieu, product marketing manager, Photo/Video Technologies, NVIDIA. "The inclusion of CUDA technology into this video processing software illustrates its broad applicability and particular value in consumer, life-style applications. We're looking forward to further collaboration and delivering the final version of the software to market."

NVIDIA first released CUDA in 2007, providing software developers with a programming environment based on the industry-standard C-language for the easy creation of applications running on NVIDIA GPUs. Numerous commercial and scientific applications have adopted CUDA technology and now consumer applications are emerging that show considerable performance improvements using the technology. Some of the first consumer applications to market are video encoding and decoding programs, which market analysts and consumer technology advocates consider prime candidates for GPU acceleration.

NVIDIA has shipped more than 80 million CUDA-enabled GPUs into the market, creating the largest installed base of general-purpose, parallel-computing processors ever produced and the latest generation of NVIDIA GeForce GPUs offer up to 240 processor cores. Processes that can be divided into multiple elements and run in parallel can be programmed to take advantage of the massive processing potential of the GPU.

The two companies plan to continue development of the software, expanding the use of CUDA within the TMPGEnc software to include acceleration of more functions and additional video formats.

For product details visit the Pegasys web site.

View at TechPowerUp Main Site

 

[btarunr]

Many Thanks to wolf2009 for sending this in.

 

[boogah]

Sweet that's y i'm still looking for a GX2 or 2 ^___~

 

[KainXS]

this alone makes me want to try to port cuda to my old 3870

so tempting

 

[Darkrealms]

Wow thats quite an increase. Isn't this the market that Larrabee was supposed to excel in?
Makes me glad I went with a GTX260, not that I do that much video editing tho.

 

[PVTCaboose1337]

I bet that I cannot afford. Nor can you. Nor can anyone.

 

[W1zzard]

The software is used to dramatically increase video decode and processing speed by as much as 446% on a GeForce GPU.

uh .. fail ? less than 5 times faster is not that much when you compare the number of flops available that the marketing teams quote us all the time? a c2d has like <10 gigaflops iirc. gtx 280 has 933 gflops. so if it can accelerate this stuff by 8000% i'd say "dramatically"

 

[soldier242]

uh .. fail ? less than 5 times faster is not that much when you compare the number of flops available that the marketing teams quote us all the time? a c2d has like <10 gigaflops iirc. gtx 280 has 933 gflops. so if it can accelerate this stuff by 8000% i'd say "dramatically"

... that really made my day, but i still like to try it out. *looks for trial version*

 

[HTC]

uh .. fail ? less than 5 times faster is not that much when you compare the number of flops available that the marketing teams quote us all the time? a c2d has like <10 gigaflops iirc. gtx 280 has 933 gflops. so if it can accelerate this stuff by 8000% i'd say "dramatically"

Apparently, there's only one flop ...

 

[WarEagleAU]

I mean this is exciting and all but it doesnt make me want to rush out and grab a gtx 260/280

 

[Darkrealms]

I mean this is exciting and all but it doesnt make me want to rush out and grab a gtx 260/280

That is true but its nice if you already had one : )

 

[DarkMatter]

uh .. fail ? less than 5 times faster is not that much when you compare the number of flops available that the marketing teams quote us all the time? a c2d has like <10 gigaflops iirc. gtx 280 has 933 gflops. so if it can accelerate this stuff by 8000% i'd say "dramatically"

I think it's a little bit higher than 10 Gflops and that those figures are per core, for a total of 40-50 Gflops on a C2Q. I'm talking about theoretical numbers of course.

In the end ~1 Tflop / ~50 Gflops = 20

IIRC Nvidia said a GPU had 20 times the power of a CPU when advertising CUDA.

 

[Mussels]

well it might not make wizzard jizz his pants, but 4.5x faster encoding is going to make a lot of people happy.

 

[Wile E]

How can they claim it's 4.5X faster without giving us a baseline? 4.5x faster than what, using what?

At 4Ghz, my cpu is pretty damn fast at encoding H.264. I somehow doubt that they mean CPUs of that caliber.

 

[tatuka666]

Not impressed yet

I was amazed by how fast Badaboom's Cuda Encoder was until I had a chance to test it myself on my 8800GT.
The quality is really bad compared to CPU based apps and there are no ways to change the settings. I then read the forums over at doom9 and they were also not at all impressed with teh quality and even concluded that a Cuda based app for transcoding video will still be SLOWER at same quality settings than CPU apps.

I hope that they will be proved wrong by newer/better software like this, but I'll wait for an independent review until I change my mind about Cuda based Apps.

 

[tkpenalty]

Instead of waste money on the nvision even why couldnt they pour that money into R&D to make better products? :shadedshu

Moreover its not a square value, and therefore 446% may be the max value. That number could have been 1337% or something if they used their resources more wisely...

 

[Mussels]

oh come on, for someone who does this already doing it 4.5x faster is going to make them a very happy camper.

hard drives, system ram etc - they're going to hold this back more than the damn GPU doing it.

 

[Wile E]

oh come on, for someone who does this already doing it 4.5x faster is going to make them a very happy camper.

hard drives, system ram etc - they're going to hold this back more than the damn GPU doing it.

4.5x faster than what, using what video card tho? I bet me or you would see a small difference, especially considering it's not unusual for me to use a lot of filters and advanced options while encoding.

 

[DarkMatter]

4.5x faster than what, using what video card tho? I bet me or you would see a small difference, especially considering it's not unusual for me to use a lot of filters and advanced options while encoding.

Oh come on guys, this is a first attempt, the second if we take Badaboom into account. Badaboom on my 8800GT it's like twice as fast than my C2Q. Sure, you can't choose anything, but it's just the first attempt, don't forget this.

Thanks God that many inventions like electricity, penicillin, explosion engines, etc, didn't have to be judged in forums. We would be still in the Stone Age, if the world was ruled by people with the mentality I'm seing on some members in the thread. :shadedshu

 

[Wile E]

I'm not questioning the value of advancements, i'm questioning their claims.

 

[Dehx]

If any of you have ever used TMPGEnc, or any of their related products, you will know that they are indeed among the slowest and worst performing of the video encoders out there. So slow and horrid in fact, that anything would merit a 400+/- percent increase in its speed. Even with that increase the software is still trash.

This news post is almost like announcing that a Boeing 747 can now hit mach 2, because they added 4 new jet engines.

 

[Wshlist]

Reminds me of ATI, ATI and cyberlink claim cyberlink's powerdirector uses the GPU to greatly increase the encoding.
But when I looked at it there's rather confusing stories, some report it's for a future update, others say it's only speeding up compressing uncompressed video not re-coding.

I wonder why ATI dropped their 'supposedly GPU but actually CPU' transcode app, it only works on X19xx cards it seems and as was revealed doesn't use the GPU 'yet' many moons ago.

So many of such promises and excellent sounding ideas never get here or stay but only flare up a moment and then fade away like some chinese CGI fireworks.

 

[Wshlist]

This news post is almost like announcing that a Boeing 747 can now hit mach 2, because they added 4 new jet engines.

You might have a point about their encoder, I don't know, but that comparison isn't working for you, a plane that could carry so many people at mach 2 would be a marvel and astounding news indeed and quite the achievement.

 

[Hayder_Master]

i want say to nvidia : "talk to someone about Unreasonable think , if he believe you this one have no mind"

 

[DarkMatter]

You might have a point about their encoder, I don't know, but that comparison isn't working for you, a plane that could carry so many people at mach 2 would be a marvel and astounding news indeed and quite the achievement.

at that plane. It would require absurdly long runways though, unless Dehx&Wshlist 747 Mach 2 ® can take-off on short runways too or maybe it's an VTOL?