Ati radeon hd 5870 - Ati News!

The first unit is indicated as Graphics Engine and it features a pipeline made of two rasterizing blocks: the new Tesselation unit, compatible with the DirectX 11 APIs and a more powerful Geometry Assembler block. The texture units were also optimized and are now capable of supplying a very high bandwidth and process over 270 billion instructions in 32-bit per second.

In order to balance this power, also the cache L1 bandwidth has been enhanced, while the Cache L2 size has been doubled (128KB for each memory controller). The memories changed little: the compatibility is still with the GDDR5 chips and the same 256-bit bus.

The biggest new features on the TeraScale 2 project are present on:

- Instruction Sets
- Stream Processor Units
- SIMD Layout
- Graphic engine
- Texture Units
- Render Back-Ends
- Display Controllers

There aren’t many relevant differences in the GPU blocks from the RV770 in an overall view. It’s also true that the previous architecture was highly enhanced to obtain a chip with great performances. Compared to the 4870, the Radeon HD 5870 have twice as many SIMD Engines, which also doubles the amount of Stream Cores and Texture Units.

Architecture

The architecture on the new ATI Radeon HD5800 was projected to manage all the features we’ve just discussed in the previous posts. Also, ATI has thought of all the games that were compatible with the previous DirectX generations. For these reasons, the TeraScale2 architecture represents what is the current “art” in the GPU world.

Thanks to the architecture, ATI claims to have doubled the performance from the previous generation, guaranteeing 2 TeraFLOPS and a more than 20 Gigapixels/s bandwidth. The same thing happened, similarly, when going from the HD 3800 to the HD 4800 series.

In this sense, AMD has worked well in offering a solution that allows the OS to see the monitors connected to the outputs as one single monitor with a bigger resolution. This was done in a driver level, so that a game, in order to support the ATI Eyefinity technology, only has to allow bigger resolutions to work.

Thanks to the driver support, it’s also possible to create group of displays that the Operating System sees as one single display. All the new features can be found on the new version of the Catalyst Control Center that was introduced by AMD last summer.

The possible uses to the ATI Eyefinity are almost endless. The displays can be configured in many ways: for example, three monitors can be placed to create a portrait or landscape mode; one can be placed in horizontal, and two in vertical; six monitors in the same row, or even in two rows of three each.

The interesting thing is that using more than one graphic card in CrossFire mode, users can connect even up to 24 monitors. But were the old limits overcome? This could be a valid question: are there support from applications and videogames or there will be a very few limited number of titles compatible with it?

The benefits available with a Radeon HD 5870 or HD 5850 are very interesting. In games, it’s possible to obtain a bigger visual that can help spotting an enemy first, or just enjoying a great panorama.

Also productivity applications such as video editing or image manipulation could benefit from it; at the same time, it would be possible to have different information available on-screen, or manage different applications at the same time: news, stock market, browser, multimedia player, word-processing applications and so on.

ATI Eyefinity

Behind the ATI Eyefinity trademark, thre’s a technology that when explained briefly, doesn’t really seem anything new: it allows users to plug more than one monitor to the graphic card and use them in different ways. Going into detail, however, we find out that there are some innovations, also thanks to the current technological scenario, that are very different from what Matrox had when they annunced their TripleHead2GO multi-monitor technology, for example.

Thanks to the ATI Eyefinity, the Radeon HD 5800 family can manage 3 monitors with the maximum resolution of 2560 x 1600 pixels for a total of 4800 x 2560, or even 6 if talking about the special version of the cards, called ATI Radeon HD 5870 Eyefinity6 Edition. We’ll see in the next articles how different this technology is from other multi-monitor technologies.

Always concerning the textures, Microsoft claims that with the DirectX 11, the decompression must be done in a completely accurate manner in order to enhance the quality. This specification is restrictive and thus helps preventing any GPU manufacturer to find a way to bypass that.

Among some of the interesting features on the DX11, we also find a new shadow management mode, called Contact Hardened Shadows, aiming to obtain a more realistic effect. The DirectX 11 also enhance the depth of field, offering a much more natural and smooth visualization.

Finally, Microsoft also implemented an Order Independent Transparency (OIT) that proves to be very efficient in the rendering of overlapping transparent objects in scenes such as those with smoke, fire, hair, foliage, fences, water and glass. Using the Direct Compute 11, the transparent pixels can be ordered in a single step.

Microsoft has also added two new formats for compressed textures, indicated with the names of BC6 or BC6H and BC7. The first one is a high-definition texture format with the compression ratio of 6:1, but not without quality losses. The textures are treated in 16-bit per channel, offering a somewhat high quality and can be used in scenes where HDR (High Dynamic Range lighting) is highly used.

The BC7 format, instead, is LDR (low dynamic range) and offers a 3:1 compression ratio with the RGB-only channels, and 4:1 when used together with the alpha channel.

Other features

Among the many new features on the DirectX 11 there’s the Dynamic Shader Linkage. Thanks to it, it’s possible to obtain a bigger flexibility of shaders, avoiding what developers tend to do nowadays when a game scenario requires too many shaders. Basically what happens now is that they create an ensemble of shaders in one single piece of code. While this approach solves some problems, it also becomes inefficient and complex to debug.

With the DirectX 11, Microsoft offers a possible solution to the problem: the usage of subroutines that allow developers to link simpler and more specialized shaders together.