One of the biggest changes of the new Radeon HD 5000 series is ATI's decision to implement a completely new nomenclature. As you may already know, Evergreen is the designated codename for all of ATI's DirectX 11 capable GPUs and instead of numbering them, ATI has given each GPU proper names. As befits the Evergreen moniker, each GPU is named after a species of coniferous trees. Incidentally, coniferous trees are those that do not lose their leaves in winter, hence Evergreen.
The Radeon HD 5800 series has been bestowed with the codename Cypress, and more specifically, the GPU found in the Radeon HD 5870 is known as Cypress XT. The Radeon HD 5850, on the other hand, is known as Cypress PRO.
Names aside, the Cypress GPU is a technological marvel that showcases Moore's Law. First and foremost, it is the first commercially available DirectX 11 capable GPU in the world. Apart from that, ATI is still the only graphics chip maker to offer GPUs built using a 40nm process.
Next, the Radeon HD 5870 and HD 5850 offer tremendous amount of computing might. Refining on the TeraScale architecture seen on the Radeon 4800 GPUs, the improved TeraScale 2 architecture on the Radeon 5800 GPUs doubles on what its predecessors had. Both Radeon HD 5800 GPUs pack over 2 billion transistors and the HD 5870 is capable of churning out nearly 3 teraFLOPS of sheer computing power. Furthermore, the Radeon HD 5870 has a whopping 1600 stream processors, 80 texture mapping units and 32 raster operator units, as we said, it is effectively double that of the Radeon HD 4870.
Raw power aside, ATI also claims to have vastly improved the Radeon 5800 series' appetite for power. Amazingly for a card packing so many transistors, the Radeon 5870 has a rated maximum TDP of only 188W, which is just about the same as the previous generation Radeon HD 4890 and HD 4870. Even more impressive however, is that it is rated to consume only 27W of power at idle.
ATI says this is achieved by more intelligent PowerPlay technologies, which throttles GPU and memory clock speeds as well as voltage according to GPU load. For instance, the GPU will have much reduced clock speeds when running your desktop application as compared to a 3D intensive game. We'll definitely be putting this claim to the test.