Radeon HD 5870 architecture features
AMD has launched the development of the Evergreen family of graphics chips. For better user perception, AMD marketers have decided to abandon the alphanumeric designations of their GPU and now all the company's new graphics products have their own names. The first representative of the new family is a chip codenamed Cypress, which should become the basis for flagship graphics accelerators with one GPU.

Later, Hemlock (hemlock - American coniferous tree), Juniper (juniper), Redwood (mahogany) and Cedar (cedar) chips will appear, which should occupy all other market niches, from inexpensive to Ultra Hi-End solutions. Thus, for about six months, AMD plans to carry out an active campaign to win our preferences.

The Cypress chip, which is the basis for the AMD Radeon HD 5870, is not just a slightly modified RV770 under a new guise. This is a completely new solution, significantly redesigned at the hardware level in comparison with the RV770/RV790. To successfully promote a new product on the global market, it must have not only high performance, but also a number of other, sometimes equally important consumer characteristics, such as functionality, high-quality visualization modes and support for modern technologies. Let's see how far AMD engineers have progressed in developing the Cypress chip and the Radeon HD 5870 graphics accelerator built on it.
Block diagram GPU Cypress

Microsoft DirectX 11 support
AMD Radeon HD 5870 is the world's first graphics accelerator with support for all functions from the DirectX 11 API set. Below is an excerpt from AMD slides that clearly demonstrate the key differences between DirectX 11 from DirectX 10 and 10.1

It is worth noting that all AMD solutions that support DX 11 are fully compatible with previous versions of DirectX. So, let's figure out what is radically new in the eleventh version.
Hardware tessellation
When applied to 3D graphics, tessellation is the process of breaking an image into smaller shapes, such as triangles or quadrangles. The use of tessellation in computer games is due to the need to increase the level of detail of three-dimensional objects. Before the advent of DirectX 11 and compatible hardware, the use of tessellation significantly loaded the memory subsystem and drives, as it required the transfer of huge amounts of data. A modern approach to tessellation should significantly reduce memory bandwidth requirements and make it possible to actively use tessellation in the latest computer games.

It must be said that the tessellation unit is present in AMD graphics accelerators, starting with the Radeon HD 2900XT, however, unfortunately, its use in the DirectX 11 environment is impossible. For tessellation in DirectX 11, additional computational stages are used - Hull Shader (surface shader) and Domain Shader (regional or zonal shader), the execution of which is impossible on accelerators of previous generations, so the existing hardware tessellation unit was not useful.

In addition to the obvious visual bonuses, it is necessary to note another pleasant fact - scalability. Let's imagine a model, the data about which is transferred for processing GPU, in particular, the tessellation unit. This unit, depending on the performance level of a particular GPU, can vary the number of object partitions to keep overall performance at an acceptable level.
Multi-threaded rendering
It is no longer a secret that one of the most effective methods for increasing the performance of computing equipment is the simultaneous processing of several data streams. The most striking example is multi-core processors, which have recently become truly accessible to the masses of consumers. Now is the time to think about more efficient use of the resources of modern GPU to speed up the rendering of 3D graphics in games. While DirectX 10 allows rendering commands to be transmitted from only one thread, DirectX 11 implements multi-threaded rendering, which makes it possible to create so-called display lists from multiple threads and execute them from the main rendering thread.
Texture compression
Texture compression methods implemented in DirectX 10 and earlier versions do not allow rendering 11D worlds with the required level of quality. That is why in DirectX 6 the developers introduced new texture compression formats - BC7 (for working with HDR textures) and BCXNUMX (narrow dynamic range RGB or RGBA textures). New methods allow game developers to use significantly larger textures, and the use of textures with a wide dynamic range will significantly improve image quality.
Eyefinity technology
Developers have been working on improving the perception of computer games for decades. Virtual reality helmets, virtual glasses and even systems for controlling characters using the power of thought - we have already gone through all this once. Unfortunately, so far none of the listed solutions can boast of mass demand. Each approach has a number of advantages, which, unfortunately, do not cover the disadvantages. With the release of the Radeon HD 5870, AMD offers its own version of expanding the boundaries of visual perception of computer games through Eyefinity technology (however, this technology can be successfully used not only in games). Let's figure out what exactly AMD offers us.

A special hardware complex allows you to connect up to six monitors to one new generation video card, and it is possible to create various connection configurations. The number and type of connectors on a specific board may vary, depending on the manufacturer's preferences.

Eyefinity technology can work with both windowed and full-screen 3D applications. Moreover, according to AMD, to support this technology in computer games it is not necessary to install any specialized patches or additional drivers. All that is needed is support for high resolutions in the game itself.
With the correct placement and selection of monitors, in games with high-resolution support, the user should be almost completely immersed in the game thanks to the so-called lateral vision.
Full screen anti-aliasing and anisotropic filtering
One of the methods for improving image quality in modern games is full-screen antialiasing (AA). Since using different AA methods requires additional effort from the 3D accelerator, maintaining a comfortable level of performance when enabling a particular anti-aliasing method is vital. New AMD HD 58xx series solutions provide almost double the performance in various MSAA (Multi Sample Antialiasing) modes compared to the previous generation HD 48xx. In addition, the user can now use the so-called supersampling, which was used at the dawn of the development of 3D accelerators, but has given way to more economical anti-aliasing methods, although it is superior in quality. The essence of this method is that the frame is rendered at a resolution higher than the set resolution in the game. Now that memory bandwidth has increased significantly, this method is experiencing a renaissance. In the future, we will have to figure out all the intricacies of modern SSAA implementation.

Another important factor is anisotropic texture filtering. According to AMD engineers, the new filtering algorithm provides ideal quality without loss of performance. At the moment, several anisotropic filtering modes are available in drivers.
GPU runs at 825 MHz and contains 1600 stream processors, twice as many as the RV770. The new GPU contains 2,1 billion transistors, which is also almost twice as many as the RV770 (956 million). The Radeon HD 5870 will be equipped with GDDR5 memory, which operates at a frequency of 1300 MHz (or 5200 MHz in DDR mode), which gives a bandwidth of 150 GB / sec. The card consumes about 180 W in 3D modes, and in idle mode - only 27 W, which is clearly less than the Radeon HD 4870 (90 W).
ATI Radeon HD 5870 Specifications
| Name | Radeon HD 5870 |
| Core | Cypress |
| Process technology (µm) | 40 |
| Transistors (millions) | 2100 |
| Core frequency | 825 |
| Memory operating frequency (DDR) | 1300 (5200QDR) |
| Bus and memory type | GDDR5 256-bit |
| Bandwidth (Gb/s) | 150 |
| Unified shader units | 1600 |
| Unified shader unit frequency | 750 |
| TMU on conveyor | 80 |
| ROP | 32 |
| Shader Model | 5.0 |
| Fill Rate (Mtex/s) | 30000 |
| DirectX | 11.0 |
| Memory Capacity | 1024/2048 |
| Interface | PCIe 2.0 |
The new ATI Radeon HD 5870 video adapter turned out to be a success: according to the results of most tests, it confidently outperformed the current top single-chip solutions (we are talking about the HD 4890 and GTX 285), and sometimes could compete with dual-chip solutions. The advantages of the HD 5870 include a very low noise level, as well as moderate power consumption, which makes it an excellent choice for a gaming PC.




