After the successful launch of the new APU generation, AMD has finally presented the cream of the crop of their product gamma - the A-series (codenamed Llano). Models from the A-series can be expected in portable PCs, all-in-one PCs, desktop configurations and similar devices which require a high level of performance, but are consumption-sensitive. The A-series APU contains three components that we’ve only been able to see separately thus far: the northbridge chip, the quad-core CPU and discreet DirectX 11 graphics, and all that at less than half the consumption of these three components when separated. Of course, packing all this onto a single piece of silicon was no easy task, making A-series the one with the largest APU surface of all Fusion platform products (228 m²) and the maximum consumption of 35 to 45 W.
As far as APU architecture is concerned, the only thing comparable enough to showcase the concept of Llano is Intel’s Sandy Bridge. Unlike Intel’s solution, where x86 cores and the cache memory make up most of the core, while no more than a third of the total resources remains reserved for graphics and multimedia, AMD took a different approach. Llano’s APU has more than 40% of the core reserved for graphics and multimedia, in order to switch (nearly) all load concerning these two categories to the DirectX 11 GPU, making processing much faster and more efficient, with a significantly lower consumption. Furthermore, this powerful a GPU has the possibility to process data over OpenCL and DirectCompute, which assures a warm welcome to this technology by graphics and parallel processing developers.
Regardless of the price range, you’ll get DirectX 11 graphics that’ll provide smooth 1080p playback and ample performance. The combination of an integrated graphics core and discreet graphics will use all advantages of the APU and the separate GPU, in terms of performance as well as consumption. AMD Turbo Core technology is present too, which enables the CPU and the GPU to dynamically adjust their frequencies in order to retain good performance while staying well within the proposed thermal frame. Of course, desktop versions will be equal to notebook ones in that they have USB 3.0 support, so that you won’t have to pay extra for third-party chips integrated in your motherboards.
One thing that’s of particular interest is that, due to the low overall power consumption, the power units don’t have to pack complex voltage filtering, which reduces the price of motherboards or enables the manufacturers to implement an extra feature or performance boost at no additional cost. This goes for PSUs as well, of course.
An important element that might give the new AMD APUs that extra edge against direct competitors is the continually improving support for OpenCL, i.e. using the power of the GPU to improve performance in demanding tasks unrelated to graphics processing, especially those that involve multimedia material manipulation. This is not just a perk of independent developers - suffice to say that OpenCL is supported by Corel, Sony, Roxio, DivX, ArcSoft, Cyberlink and many others. Seems much more serious now, doesn’t it?
With the alleged, much lower desktop “kit” (motherboard + CPU) prices, barely going over the barrier of 200€, even with top models, all this seems very attractive, but only provided that the performance matches the levels promised by AMD.
The direct gains enabled by the Brilliant HD technology will be felt by users in several fields. One of the interesting features is AMD Steady Video, which dynamically processes all video played on the user’s PC in order to stabilise the shot and eliminate all undesirable side effects made by the recording person shaking the camera. This is especially relevant for informal, home-made videos frequently uploaded to social networks, mostly of amateur quality and containing the said problems. Even when streaming content from YouTube or similar sites, there will be an AMD Steady Video pop-up, with a button to (de)activate the use of this technology. This enables the user to retain the shaky effects in scenarios where they are intentional.
Furthermore, the APU will enable CrossFireX capabilities by combining the integrated with the discreet graphics, which can increase performance significantly on portable PCs. Naturally, this doesn’t exclude the option of keeping only one of these active, to balance out performance and power consumption.
The existence of a unit capable of parallel information processing (GPU) and its use in various fields is one of this APU’s key features. Fusion Compute Link is an internal bus four times the speed of PCI Express 2.0, which connects parts of the Fusion AMP directly, making the process of transferring data back and forth between the CPU and the GPU four times faster. GPU computing has been improved as well by adding a chart that memorises the placement of data inside the memory, which in turn eliminates the need for the same data to be placed in both the CPU and the GPU memory.
USB 3.0 has finally become an integral part of the Fusion controller hub, relieving the PCI Express bus, as was the case with a separate controller chip, which should theoretically improve transfer speeds by some 10%.
As with NVIDIA’s CUDA technology, applications using OpenCL have significant performance gains, comparable to much stronger and more expensive CPUs; this turns notebooks into very powerful workstations that can process complex tasks, especially multimedia ones.
The technology named AllDay Power by AMD not only provides good autonomy owing to the increased platform efficiency, but also reduces heating to the minimum; AMD has been praising their ability to eliminate hot spots from notebooks, even on the bottom, where there’s traditionally the most heat. The autonomy itself was fantastically improved, being increased by almost 50% compared to last year’s Danube platform, in both idle and full load modes, which is allegedly better than anything Intel’s Sandy Bridge mobile CPUs can offer.
