Pegasys Inc. (Ryogoku, Sumida-ku, Tokyo; CEO: Tak
EBINE) today announced at
the NVISION®
2008 visual computing conference, being held August 25th to 27th
in San José California, that it will use NVIDIA® CUDA™
technology to offload much of the processing required for TMPGEnc 4.0 XPress
from the CPU to the graphics processing unit (GPU). A beta version of the
software shows up to 446% performance increase over the CPU-only version.
What is CUDA technology?
NVIDIA’s CUDA technology is the world’s only C
language environment for the GPU. It enables programmers and developers to
create software for quickly solving complicated computation problems using GPU
multicore parallel processing capabilities. To date, NVIDIA has shipped over 80
million CUDA-enabled GeForce® 8 Series and higher GPUs, the largest
installed base of general-purpose, parallel-computing processors ever created.
The latest generation of NVIDIA GeForce GPUs offer up to 240 processor cores,
compared to a maximum of the four cores found on the highest-end CPU. Any
process 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.
CUDA accelerates filter processing in beta version of
TMPGEnc 4.0 XPress
Pegasys has incorporated CUDA technology for filter
processing into the new beta version of its flagship video encoding software,
TMPGEnc 4.0 XPress. “In a very short time we have been able to use CUDA
technology to offload decoding onto the GPU and we look forward to working with
NVIDIA to expand it to the encoding, as well. Adding GPU support into the
filter processing was the best way to see a dramatic result in a relatively
short time.” (Comment by Tak EBINE, CEO)
“NVIDIA CUDA technology has already made a
substantial difference in the performance of TMPGEnc, a very popular creative
video editing tool for consumers, yet we are just scratching the surface
of the visual computing capabilities of NVIDIA GPUs. As Pegasys continues to
offload more and more of the functionality to the GPU, we will see dramatic
results that will be a boon for both video enthusiasts and casual home movie
users looking for simple and fast ways to edit, share, and enjoy their
video,” said Michael Steele, general manager of visual consumer solutions,
NVIDIA.
Test results show impressive performance gains
In the beta version, application performance has been
increased by up to 446% with CUDA technology, compared to the Intel Core2Quad
Q9450, a top-of-the-line CPU. This demonstrates that CUDA technology will have
a major impact on encoding speed, which requires a lot of filter processing.
System Specification
CPU
C2Q Q9450 (2.66 GHz)
Memory
2 Gbyte
Graphic
GeForce GTX 260
OS
Windows XP SP2
TMPGEnc 4.0 XPress
Ver. 4.6.0.257
Test Contents
Encoding HDV file to
MPEG2 for DVD-video format for 3 minutes
Test１
De-interlacing + resizing
Test２
Test１+ 2D & 3D noise reduction
Test３
Test 2 + sharpening edges
Test４
Test 3 + smartsharp
Test５
Test 4 + color correction
Result (Frame Rate)
Test
CUDA Execution (Beta)
CPU-only
CUDA acceleration
Test１
2,803
5,253
53.36 %
Test２
1,377
1,667
82.60 %
Test３
1,217
576
211.28 %
Test４
993
224
443.30 %
Test５
928
208
446.15 %
Release of products with CUDA technology
The beta version of TMPGEnc using CUDA technology is
available from Pegasys upon request. Release dates for final versions of
TMPGEnc using CUDA technology, starting with TMPGEnc 4.0 XPress, have not been
announced.
