Ati Radeon HD 4800 Series: description of architectural features of the outdated series of video cards

Ati Radeon HD 4800 Series именуется классически. Series Ati Radeon HD 4800 Series is called classically. The first digit indicates the adapter generation. The second hides the family. The latter two point to a specific model. Initially, the cost of video cards of this generation began with a mark of 5000 rubles. Now, when the release of these devices has ceased, the secondary market can offer similar devices at a price of 1,000 rubles.

Core

The video processor was named RV770. It includes more than 900 million transistors. Of course, this is quite a lot, but not a record. However, the improved 55-nm process technology allowed to reduce the price of their release and briefly bypass competitors.

ATI Radeon HD 4800 Series довольно сильно снижены по сравнению с ядром предыдущего поколения. I must say that the frequencies of the GPU ATI Radeon HD 4800 Series are pretty much reduced in comparison with the core of the previous generation. However, the increased power of texture processors should balance the lack of frequencies of the GPU.

Special attention should be paid to the memory of GDDR5. Its application made it possible to raise the bar of the bus bandwidth without increasing its bit capacity. Of course, the price of GDDR5 is high, but it plays a minimal role, as it simplifies the location of all devices on the board.

The most problematic place for video cards with serial number 3 is texture processors. Increasing their number to 40 allowed to solve the main problem.

ALU Improvements

видеокарта ATI Radeon HD 4800 Series очень зависима от уровня оптимизации кода шейдеров в 3D-программе. Any ATI Radeon HD 4800 Series graphics card is very dependent on the level of code optimization for shaders in the 3D program. However, the adapters of the series are equipped with eighty ALUs, and on each core of the processor there is a task manager for optimizing work in automatic mode. In the worst case, the code will be processed simultaneously in 160 ALUs. If each computational unit is loaded to the maximum, the computing power will be equal to one teraflop. This indicator is comparable with the ninth-generation NVIDIA flagship.

The End of the Ring Architecture

ATI Radeon HD 4800 Series под микроскопом, можно увидеть, что в общем сохранены все технические особенности предыдущего поколения ядра. Considering the ATI Radeon HD 4800 Series video card under a microscope, you can see that in general all the technical features of the previous generation of the kernel have been retained. However, the architecture is now non-colloidal.

Memory controllers are united by interfaces that operate bidirectionally. Blocks that do not need quick access to instructions are located around the kernel. In addition, next to the processor is a switch, it connects devices for which the bandwidth is not critical. If the efficiency of the work of the elements of the core of the last generation could reach 85%, then the improvements made it possible to bring it to the practical maximum. ATI specialists say that the computing units have become 40% more efficient.

Dispatch Processor

The task manager can be called the key device of any graphics processor. His job is to effectively allocate resources. In this case, all parts of the kernel must be loaded in the same way. If this condition is met, the performance will be maximum.

количество блоков для хранения и обработки кода увеличилось. In the ATI Radeon HD 4800 Series handler, the number of blocks for storing and processing the code has increased. Also, the number of SIMD-arrays became larger. Due to this, each array now works not in isolation, if necessary, data exchange is possible. In addition, the algorithms of the task manager have also been optimized.

Texture blocks

The third generation of video adapters had only 16 blocks of textures, which is absolutely inadequate for most 3D applications. And for each TMU there was only one filtering unit. This was especially noticeable when anisotropic filtering was enabled , but it was used everywhere.

The developers took into account the experience of previous mistakes. Now in each TMU module there are 16 blocks that perform texture sampling, and 4 blocks for filtering and addressing. Coupled with the increased throughput of TMU, the performance of this element has increased by 2.5 times.

Modules, like in the older generation video adapters, are grouped by 4. However, due to optimization, the number of transistors here has decreased.

Cache

характеристики которой заметно выросли по сравнению с младшими адаптерами, получила новую систему кэшей. The ATI Radeon HD 4800 Series video card, whose characteristics have grown significantly compared to the smaller adapters, received a new cache system. Now the data exchange with the first level cache occurs at a speed of about 480 Gb / s, the second - more than 380 Gb / s. In addition, all SIMD cores are now installed with their own memory, which allows you to operate simultaneously with a large number of information.

New opportunities

In addition to the basic innovations, it is not superfluous to note smaller ones. All HD 4800 graphics cards have Realtek controller in their kit, which allows to work simultaneously with eight channels. The maximum output bitrate is just over 6 Mb / s. The upper sampling rate is at 192 kHz.

Advanced features are very important for those who are going to connect a multimedia system to the computer. о которой и без того лестны производителю, из-за описанных выше особенностей получила в свою армию фанатов еще и пользователей, не интересующихся компьютерными играми. ATI Radeon HD 4800 series, reviews of which are already flattering to the manufacturer, because of the features described above, has received in its army of fans also users who are not interested in computer games.

Built-in universal video decoder not only easily copes with the playback of high-definition video, but can simultaneously display it on two different monitors. In this case, different signals can be output to each output device.