AMD Radeon HD 7970

Present a short preview of the features of the Radeon HD 7970, the only one in which can reveal details in advance of the moves with which AMD will change its offer to the market in early 2012.

At the close of 2010, with the launch of the Northern Islands and the family of Radeon HD 6900, AMD introduced a first ex-design with the past with Southern Islands the distance is even more pronounced because of the big model of the architecture of the heart. After a year of stagnation for the entire field of 3D graphics, especially in the niche of products dedicated to high-end gamers and more advanced users and enthusiasts, the news that AMD has in place are really a lot.

Start from the production process with the jump from 40 to 28 nanometers that has allowed to cram into the silicon of the Radeon HD 7970 Tahiti well 4.312 billion transistors. This production technology allows the graphics processor to operate at a frequency of 925 MHz while still leaving considerable room for growth, margin will be just that the playing field on which the partners of AMD will be able to indulge themselves with the factory overclocking, so as to offer the avid gamer products higher performance than standard.In terms of memory, suggest the introduction of the controller with a 384-bit wide bus (six-channel 64-bit) supports five GDDR5 memory operating frequency of 5.5 GHz equivalent to a theoretical maximum bandwidth of 264 Gbytes / s.

Full of new architecture
With Southern Islands AMD introduces a new generation of architecture identified as self-explanatory as the Graphics Core Next present the overall scheme referring to the reading test in the next issue of PC Professional for an analysis in detail. The core architecture has 32 drive bays, each of which integrates 64 stream processors for a total of 2,048. Right here w find one of the most interesting because the designers have abandoned the approach AMD Vliw5 and Vliw4, used respectively for the Radeon HD 6800 series to 6900 and for those in favor of something more flexible and less structured. A Compute Unit is organized into 4 units inside SIMD ( Single Instruction Multiple Data ), each of which contains 16 stream processors, in turn, operated by a single programmable scheduler. Alongside the SIMD units are 4 texture units and a structure of registers and cache completely redesigned for high performance not only with typical workloads of video game graphics engines, but also with multimedia applications, computing, productivity and scientific.

Performance, but always under control consumption
Among the many new features stand out the new control technologies and energy efficiency related to the placing of double-flow performance, a product of this caliber, is the first goal in the ladder of priorities. Hence that behind names such as AMD and Powertune ZeroCore lie Power monitoring and control of power supply voltages, currents, consumption, frequency and temperature without interruption while the GPU is active. The operating frequency is handled with extreme granularity so as to offer maximum performance within the parameters of maximum consumption. When the control system detects a situation of static image rather than being idle at preparation of the system here is the emerging energy-saving solutions: from 24 watts down to less than 3 watts thanks to the effective shutdown of the GPU, the maintenance of the circuit only refresh the memory controller and PCI Express. All this not only when you have one card in the PC, but also in the case of Crossfire configurations.