It is 7 years late, but it’s finally here: 10 times faster using the APU

Apr 24, 2012 14:39 GMT  ·  By

Today, AMD announced on their official blog that their collaboration with Adobe to bring the first implementation of OpenCL heterogeneous compute within the Adobe Creative Suites family has brought up to ten fold performance improvement in Adobe Photoshop CS6 and Premiere Pro CS6.

AMD has promised computer users and the industry altogether that the fruits of the ATI buyout will come very soon. That “very soon” seems to be defined as 7 years in AMD’s dictionary.

AMD has long been a company with not much interest in software work. Every time they released a new product, they focused on making the best hardware they can possibly make, and software optimization was left for anybody else to do, except AMD themselves.

Recently, they decided to lessen the work on drivers for their HD 2xxx, HD 3xxx and HD 4xxx series, as most of the bugs have been ironed out.

The problem is that all these GPUs have features that haven’t been enabled yet because of the lack of software work.

Two such features are the Tessellation unit and any kind of GPU compute application. We were emphasizing on how important such an application would have been two years ago in our AMD drivers news update.

Building an application that uses the Stream Processing Units (SPU) inside these “old” GPUs would have been a very useful proof of concept. That would have encouraged developers to build similar applications and it would have also been a very good marketing point.

When the first Fusion CPUs launched, there was absolutely no user-friendly application that could harness the power of the SPUs inside the Fusion iGPU.

Considering that all current Fusion CPUs use the same VLIW architecture that the HD 2xxx through HD 4xxx series are built upon, AMD would have had a lot of different applications similar or better than Intel’s QuickSync.

Therefore, right now, it’s great that AMD can accelerate Adobe’s Blur Gallery and Liquify features over 10 times by using the integrated GPU in the new “Trinity” processors.