After AMD covered $ 549 and $ 449 in Tahiti-based Radeon HD 7970 and HD 7950 in January, AMD is bringing its latest-generation GPU into mainstream brackets today. The new Radeon HD 7770 and 7750 use the same 28nm design process and GCN (Graphics Core Next) architecture as the flagship 7000 series, but in cheaper configurations.
By switching to 28 nm, AMD can press 1,500 million transistors into a 123 mm2 chip. In addition, the HD 7700 series chip is 26% smaller than the HD 6770 and contains 44% more transistors. As impressive as these numbers are, gamers will be more pleased with AMD's pricing: The HD 7770 is priced at $ 159 – in line with the GeForce GTX 560 – and the HD 7750 is even cheaper at $ 109 fights the GTX 550 Ti.
AMD Graphics Core Next
As explained in our preliminary review of the Radeon HD 7000, the new series is the most significant overhaul of AMD's graphics architecture in the last decade. At that time, AMD took over the Graphics Parallel Core architecture and used groups of scalar processors that process very long instruction words, commonly abbreviated as VLIW. Radeon HD 5000 cards used VLIW5 and last year's HD 6000 series was upgraded to a more sophisticated VLIW4 architecture. However, the HD 7970 and other cards based on the Tahiti core are replacing VLIW stream processor clusters with Graphics Core Next processing units.
GCNs are basically GPUs that can do both graphing and computing tasks with great efficiency. Sound familiar? This is AMD's answer to the Nvidia Fermi architecture that they switched to in 2010. Essentially, a shader cluster is now called a GCN computing unit, and each unit is a superscalar processor with scalar and vector elements that follow a new non-VLIW instruction-set architecture. This architecture is more efficient and delivers more performance per millimeter square of GPU chip area.
The HD 7770 and HD 7750 feature a new 9th generation tessellation geometry engine with optimizations such as increased vertex reuse, improvements in off-chip buffering and larger parameter caches. This helps to increase the performance for all tessellation factors with up to 4 times the throughput of the HD 6900 series (Gen 8).