14 DAY TRIAL // We talked a lot about AMD’s lack of software optimization right before Trinity’s official launch. AMD needs to work on software development, as relying on Intel’s compilers is a very bad idea.
CyberLink software company has just updated its media processing software suite with complete support for AMD’s Video Codec Engine (VCE), as announced on its own official website.
Every GPU and the iGPUs inside APUs have a lot of computing power hidden inside. We got used to see 3D applications using that power, but fast floating point performance is a desirable feature all over the board.
AMD’s Trinity has a lot of potential and has already proven to be much more efficient than Intel’s Ivy Bridge, and it displayed considerably better gaming performance.
The problem is that the company’s new APU has rather mediocre x86 performance, due to the lack of FPU computing power.
This was actually an intentional move supported and perpetuated through AMD’s latest designs by its former CEO, Dirk Meyer. He bet on the fact that only integer performance needed to be increased, as AMD already had a lot of FPU power coming from their GPUs.
While the bet is valid, it is very hard to put in practice and a lot of software work is required to ensure that programs really know how to tap into that iGPU power.
CyberLink states that a 360% performance improvement is obvious in PowerDirector 10, when GPU acceleration is activated.
The performance improvements are not always this impressive, but generally, if the improvement achieved is more than 30%, it shows that going after the GPU FPU power was the right thing to do. It would have been incredible, if not impossible, to get this amount of power by upgrading the x86 FPU.
Such performance increases haven’t been experienced ever since AMD’s K7 Athlon launch back in 1999, when the K7 CPU was over 50% faster than anything Intel had on the market at the time.