The R-Series architecture (Fig. 2) delivers twice the graphics performance of comparable discrete graphics platforms in an integrated system. The APU is capable of driving four displays by itself and supports AMD's Eyefinity technology that allows displays to be merged into a single logical surface. The chips target the embedded market with at least a 5 year availability and dedicated embedded support. The series targets mid to high end, graphics intensive applications like digital signage, medical imaging, point-of-sale (POS) and casino gaming.
The x86 cores incorporate AMD's Turbo CORE 3 Technology that automatically balances performance and power consumption. The chips support a 4 Mbyte L2 cache and 1600MHz DDR3 DRAM. This includes DDR3 support for 1.5V/1.35V/1.25V DIMMs.
AMD's virtualization technology (AMD-VTM) supports SVM pause count capability and SVM disable and lock support. It has improved world-switch performance and it supports rapid virtualization indexing.
The APU Thermal Design Power (TDP) ranges from 17W to 35W with an average power below 13W. The x86 cores are based on AMD's 2nd generation “Bulldozer”-based core architecture (Fig. 3). A pair of cores share a floating point unit allowing better floating point performance and system utilization. The floating point unit can operate as a single 256-bit unit or a pair of 128-bit units.
The APU's graphics support handles a range of output solutions. It can drive a single 4k by 2k display at 60Hz. It can also drive four independent displays. The AMD Eyefinity technology allows these displays to be combined as a single surface and this can be extended to more displays with the addition of an off-chip Radeon controller via AMD Radeon Dual Graphics support. The APU graphics interface supports DisplayPort 1.2 with 5.4 Gbit/s link speeds as well as daisy chaining with HDCP 1.4 support.
The GPU has Microsoft Windows' Enhanced DirectX 11 Graphics Multimedia support and Shader Model 5 support. It supports HD decode of H.264, VC-1, MPEG-2, and DivX. The hardware encode support for 1080p H.264 handles 60 frames/s.
The GPU supports OpenGL and OpenCL (see Multicore Mobile GPU Handles Computation Chores). The latter allows the GPU to handle processing chores in addition to driving displays.
The APU supports a range of security features including AMD's Secure Asset Management Unit and encryption instructions. These help lower the power and CPU overhead required when dealing with protected content.
The top end, FS1r2 R-Series APU supports a x16 PCI Express Gen 2 interface and fits into a 35mm by 35mm, 722-pin PGA chip. The FP2 in a smaller 27mm by 31mm BGA package supports x8 a x16 PCI Express Gen 2 interface. Each has an additional 4x1 or 1x4 Gen2 PCI Express interface.
The R-Series works with the AMD A70M and A75 hub controllers. These add interface support for peripherals such as gigabit Ethernet, USB 3.0 and SATA II. The SATA controllers have RAID support. The A75 has legacy PCI bus support.
The R-Series represents a major step forward in system performance and integration. I had a chance to talk with some of the AMD designers about the R-Series architecture including Jon Lotz, Senior Fellow at AMD's Mile High Design Center in Fort Collins, CO. Below are some videos that can be found on Engineering TV.
Kevin Hurd provided a more detailed description of the improved computational performance found in the new R-Series APU.
If you want to see a 9 screen system driven by an R-Series APU then check this out. Kelly Gillilan explains how the APU is combined with a Radeon graphics card. The system can actually support 10 screens.
Kelly also talks about the performance differences between standalone graphics as well as APU and APU augmented graphics.
Of course, if you want to find out about the difference between AMD's G-Series and new R-Series APUs we have that too. I spoke with Kelly Gillilan from AMD about the completely new core design as well as the significant performance improvements they added.
Finally we get to play some slots with an R-Series demo.