Review: GPU-accelerated video-encoding. Who is better, AMD or NVIDIA?

by James Smith on 28 October 2009, 12:09

Tags: AMD (NYSE:AMD), NVIDIA (NASDAQ:NVDA), HiS Graphics, BFG Technologies

Quick Link: HEXUS.net/qaukz

Add to My Vault: x

A brief trip down History Lane

Before the release of DirectX 10 compliant hardware, the limitations imposed by DirectX 9 hardware meant that programming the GPU directly, as well as adapting existing code to work on the GPUs' architecture, was an onerous undertaking.

As a result, the only mainstream application that existed had been developed in conjunction with ATI, for its X1950, X1900, X1800, X1650, and X1600-series cards. This took the form of a GPU-accelerated Folding@home client back in September 2006.

Later that year, AMD announced collaboration with RapidMind and released its Close to Metal (CTM) API for RapidMind’s development environment, as well as launching the first of its stream computing-based products. However, this didn’t result in an emergence of GPU-accelerated consumer applications, most likely due to the focus being on accelerating enterprise-level programs.

NVIDIA

That same month and on the same day as the GeForce 8800 GTX was launched, NVIDIA announced its own stream computing initiative, known as CUDA. As with AMD, NVIDA initially pitched the GPU-accelerated development tools for use in the enterprise environment.

Despite the availability of CTM, users of AMD’s Radeon HD 2000-series cards were left without any mainstream GPU-accelerated applications for almost a year - the Folding@home GPU1 client would only work with X1000 series cards. Radeon HD 3000-series users faired a bit better; they only had to wait six months.

2008

In June 2008, a couple of months after the release of the GPU2 Folding@home client for ATI, NVIDIA 8-series owners finally got a taste of what CUDA could do for them outside of playing games, with the release of the GPU2 Folding@home client.

In August 2008, NVIDIA launched the first of three downloadable demo packs and patches titled “Graphics Plus Power Pack #1”. The main feature launched along with the power pack was PhysX - GPU acceleration for in-game physics. More interesting, though, was a demo of the CUDA-accelerated video-transcoding application known as Badaboom! 1.0.

Four months later, ATI launched its stream initiative and associated drivers for the HD 2000, 3000, and 4000-series cards. At the same time it also launched a new version of its Avivo Video Converter. However the initial version was only compatible with HD 4800 and 4600-series GPUs, and wouldn’t work with GPU acceleration under Vista x64. It wasn’t until a significant chunk of 2009 had flown by that these issues were addressed. The latest version now works correctly under Vista x64 and supports HD 2000, and HD 3000 series cards, as well.

Also during December 2008, NVIDIA launched its second power pack. Of most interest were trial versions of the latest build of TMPGENC Xpress, in addition to CyberLink Power Director 7.0, both of which added GPU-accelerated video effects. At the same time, although not included in this power pack, a CUDA-accelerated version of Seti@home was being rolled out to people already signed into the Boinc distributed computing network.