2013年11月30日星期六

ATI

ATI (Array Technology Industry) is a world-renowned graphics chip manufacturer NVIDIA par, the Chinese name of "the rule of the day. Between 1985-2006 is an important global graphics chip company, headquartered in Markham, Ontario, Canada, to become a part of the company until 2006 AMD acquired by the company at $ 5.4 billion heavily. ATI has more than 3,700 employees in the Americas, Europe and Asia, and a turnover of $ 2.2 billion, is specialized in the design and marketing of PC display card, graphics processors, chipsets, set-top boxes, digital TV, Electronic game consoles and hand-held devices such as fabless semiconductor company. founder of

founder the Guoyuan

ATI's when he was CEO of Guoyuan [1] (Kwok Yuen Ho) 1950 was born in a once wealthy Guangzhou family then left the Chinese mainland. A major turning point of his life in 1974, graduated from Taiwan's National Cheng Kung University, and received a bachelor's degree in electronic engineering. Then he worked for the data control system, Philips Electronics, and National Semiconductor Corporation, and then later on Ren Wangshi Electronics Company, Ltd., a PC manufacturing and assembly plant.

1984, Guoyuan immigrants to Canada. Unlucky, although he has 10 years of work experience and a degree in electrical engineering, but it is still difficult to find a match and his job. However, the the Guoyuan job market is not the only losers. In 1985, he and two other Hong Kong immigrants, Benny Lau (VP of Product Development, has been retired) and LEE KA of LAU (Liu Li Kai, strategy plans, vice president), and jointly create a their own company, also is famous of Array Technology Industry .

July 24, 2006, ATI, AMD acquired $ 5.4 billion heavily. As a sub-brand for its graphics card business. Until August 2010, released the Radeon HD 6000 series graphics cards at the same time, AMD renounced ATI brand [1].

Edit this paragraph AMD-ATI platform business PC platform for business

as the company's core business, AMD-ATI with its leading technology in the PC platform for PC and workstation users top graphics performance. Using the most advanced semiconductor manufacturing technology, production of AMD-ATI Radeon series graphics processor will give PC users a the top visual experience. Mobility Radeon graphics processors, AMD-ATI produced specifically for mobile products in the notebook market in 2009 has occupied more than half of the share reached 57% share of the mobile graphics, this graphics processor can not reduce the performance of the machine processing power and mobility, provide more exciting features.

2010 AMD-ATI GPU market share in the second quarter beyond NVIDIA

integrated graphics counted, Intel is still the dominant graphics chip field, the share of the second quarter was 54.3%, AMD 24.5% and NVIDIA ranked in the following 19.8%. The ranking also reflects the conversion of the two position of AMD/NVIDIA. If you have only discrete graphics share in discrete graphics market share in the second quarter of 2010, AMD was 51%, and 49% of the more than NVIDIA's just good. Only 2% of the gap, but the completion of twelve qualitative change in the conversion of the market share. (Mainly by NVIDIA Fermi architecture design yield theoretical design beyond the existing production technology thus resulting in insufficient supply) to know, NVIDIA's share in the year ago is still ahead of the AMD-ATI course also not a monopoly. The PC market is not static graphics chip, AMD and Nvidia share uncertain. The AMD share the upper hand, the next quarter or next year Nvidia may prevail, but to compare the relative INTEL strong monopoly in the CPU field repression AMD is not really a contest between the AMD-ATI and Nvidia is one-sided The two sides each outcome.

ATI the integrated graphics products, the value and mainstream users and his ilk can be affordable desktop and PC prices for high-performance graphics processing functions, cost-effective graphics.

edit this paragraph Product Development ATI positioned himself as a computer graphics company was established in August 20, 1985, the same year in October ATI ASIC technology developed its first graphics chip and graphics card, after mainly involved in OEM (Original Equipment Manufacturer) business, manufacturing graphics chips for large PC manufacturers (such as IBM). The PC only IBM and other competitors machine sales, but valuable, including all accessories are not alone retail software, so the OEM is the only way. the family of

EGA Wonder and VGA Wonder graphics card

For the sake of the future of the company, ATI decided not to mass manufacturing basic 2D graphics chip. In 1987, ATI sale EGA Wonder and VGA Wonder family of graphics cards. These graphics capabilities than IBM's own display device EGA/VGA Wonder is a can be used for any kind of graphical interface software and display a single card on the market, provide more high-speed image as a traditional PC, thereby attracted the attention of PC manufacturers and users! transition to

Radeon 9700

Rage Radeon100 small test chopper to the Radeon200 rise last to Radeon300, brilliant, which has gone through numerous ups and downs. Matrox Parhelia-512 seems to grab the the DirectX 9.0 top spot, but the earliest is not necessarily the best, ATI Radeon 9700 have chosen a better time. Past ATI product development is always a step slower than the nVIDIA like 99 years before AMD. If you want to beat the competition, you have to fight on in the product development cycle time, choose a short silence, then the thin plot thick hair. AMD's K7 doing just that, and ATI's Radeon 9700 is obviously also a classic case. The the

first launched Radeon 9700/Pro can be described as refreshing, not only the first to achieve the DirectX 9.0 hardware support, integrates 110 million transistors for the first time ever! As a new generation of products, Radeon 9700/Pro achieve 256Bit bit wide memory interface to achieve this value. From memory width bit rendering pipelines, texture mapping unit indicators concern, Radeon 9700/Pro the performance is impeccable. Compared with the GeForce4 Ti, Radeon 9700/Pro completely beyond it, and can not be counted as a product of the same era. Can be responsible for that, then on the progress of the product development, ATI nVIDIA beyond first. Radeon 9700/Pro configured with four programmable vertex described King pipeline, but also improve the polygon set engine can provide support Vertex Shader 2.0, Pixel Shader 2.0, NURBS, Displacement Mapping technologies such as the most perfect. Full-screen anti-aliasing technology has always been continuous efforts in the direction of the GPU manufacturers, while the Radeon 9700/Pro bring the industry is Smoothvision 2.0, with the new generation of HyperZIII memory compression technology directly led to greatly enhance the practical value of the full-screen anti-aliasing.

Radeon 9500/Pro

There is no doubt that the high price of the Radeon 9700/Pro only ATI's banner the truly crazy Radeon 9500/Pro. The Radeon 9500/Pro Radeon 9700/Pro R300 GPU, merely reduced to 128Bit DDR memory interface. Relatively speaking, the Radeon 9500 speed reduced some more, because it has only 4 pixel rendering pipeline. But even more surprising is that the many Radeon 9500/Pro can be modified via software Radeon 9700/Pro, performance improved significantly at this time much fans respected. Perhaps see the Radeon 9500/Pro small flaws, perhaps in order to reduce costs, ATI launched quickly based on RV350 kernel the Radeon 9600/Pro of. Radeon 9600/Pro same full support for DirectX 9.0, but only configure two programmable vertex described King pipeline, and reduced to four pixel rendering pipeline, only 128Bit DDR memory interface at the same time. In this way, Radeon 9600/Pro hand lost light, so the popularity of speed is not fast. Of course, derived later ATI Radeon 9550 or very successful, many ATI Fans hailed as a classic of classics.

Radeon 9600

contrast the GeForce FX5950U Radeon 9800XT, we have been able to find a strange scene. In most DirectX 8.1-based games, the gap between the two is minimal, and DirectX 9.0 games, Radeon 9800XT has obvious advantages. Moreover, ATI's Radeon 9600 series also has a similar advantage, so nVIDIA GeForce 5700/5600 series will be seriously affected. Announced the news from the official point of view, GeForce FX5950U Radeon 9800XT DirectX 9.0 implementation, there are some differences, and even cause performance gap. Although both support Pixel Shader 2.0 and Vertex Shader 2.0, but the rendering accuracy, the number of instructions, Shader length is not the same. The amazing

Radeon 9800 DirectX9 3D dynamic mapping technology

of original nVIDIA GeForce FX relative to one of the main advantages of the Radeon 9700Pro is that we can perform the Pixel Shader procedures lengths up to 1024 instructions, but ATI subsequent release of the Radeon 9800 series can execute arbitrary commands on the length of Pixel Shader procedures than the GeForce FX5950 into step. It's all thanks to ATI's F-buffer technology is mainly used for storage in the rendering process intermediate results, thus avoiding all the pixels are written to the frame buffer to improve work efficiency. the growing of

nVIDIA with ATI the confrontation

nVIDIA forces, as well as in the development of the interface monopolistic behavior that Microsoft is very angry, even once rumored the nVIDIA exit DirectX 9.0 development group. The saying goes, no wind, no waves, which at least proved nVIDIA relationship with Microsoft as before. As we all know, the Pixel Shader and Vertex Shader only the D3D interface under application, exit the the OpenGL organization's Microsoft bent completely eliminated OpenGL interface, and has been largely achieved. In this case, nVIDIA situation naturally very negative, very likely in the face of a new round of Version 3.0 Pixel Shader and Vertex Shader competitive disadvantage. Even more strange is the ATI R & D team originally responsible for developing DirectX standard before Microsoft members, but also for the the nVIDIA most embarrassing. Comparison to the same level of graphics, nVIDIA OpenGL performance has a great advantage, and the the ATI D3D speed even better. The reason for this situation is not a purely technical reasons, "co-operation" between the vendors and the "containment," played a decisive role.

ATI and nVIDIA first DirectX9 against a conclusion, then ATI can be said that victory. However, be sure not to underestimate the the nVIDIA ambition ATI eating Radeon 9550 stakes, nVIDIA PCI Express Raiders have already started to and on this ATI into the abyss. The recovery. Kiskiskis

SM 3.0 the defeat

ATI well other GPU manufacturers nVIDIA has sounded the alarm, Radeon 9700/9800 is to let the nVIDIA know backwardness feeling. The 4 × 2 pipeline architecture of the GeForce FX5800 128bit GDDR2 memory is making nVIDIA in the the performance throne contention completely lost ATI, for nVIDIA terms, make it real restless also more than ATI's high-end products. Radeon 9550 and the derivative to the X300-X550-X600 PCI Express interface group fit for market share erosion was very serious.

In this case, expect FX5200/5500/5600/5700 ilk to restore the situation is not likely to exist the problem of high cost and nVIDIA bridge chip. How to completely suppress the ATI, the imperative which became nVIDIA. Fact, nVIDIA has a slight lead in product development progress at ATI. Before the release of GeForce 5900, GeForce 6 Series has entered the stage of research and development. nVIDIA GeForce 6 inputs engineering R & D personnel no less than 500 people, the number of R & D spending billions of dollars. Since so far, can be completely declared nVIDIA GeForce 6 product line layout completed NV3X core gradually withdraw from the market. GeForce 6200TC against nVIDIA mortal malady is called integrated graphics, the ATI X300-X550-X600 camp completely suppress competition, coupled with high-end SLI technology with GeForce 6600GT GeForce 6600LE series to contain, it should be said nVIDIA odds .

But from a technical point of view, X300-X550-X600 camp had the face of a mediocre reality. Today, more and more games began to support Shader Model 3.0, including the "Splinter Cell" and "Far Cry" and "Age of Empires 3", "Call of Duty". GeForce 6200 and GeForce 6600LE full support for Shader Model 3.0, ATI graphics card situation is clearly lagging behind in this regard.

Shader Model 3.0

Shader Model 3.0 the game development greatly enriched programming model, to convenient simple game developers do a better game. Shader Model 3.0 is applied to environmental surfaces and mixed specular light source. , Shader Model 3.0 and Shader Model 2.0 compared to the biggest advantage is that the technology has a displacement map, many complex lighting algorithms can not be achieved in Shader Model 1.1 and Shader Model 2.0. Kiskiskis

for any manufacturers, many of the market is not always abandon frontier. At that time, ATI is hoping X300-X550-X600 camp be able to stick to the low-end market, while the X700 is at least a threat to the GeForce 6600 series. But now it seems, ATI's wishful thinking is clearly unrealistic. Due to backward technology research and development, ATI has lost in the second round of the the DirectX9 competition arena. Of course, the failure of the X700 also means the defeat of the entire product line. X800 the subsequently derived X850 series, but in the high-end market simply can not open up the situation, it is difficult confrontation with the GeForce 6800 series front.

X1600

a pipeline built more than one unit of work is not a new technology, the GeForce4 Ti previous Radeon 8500 graphics are using these technologies. And rely solely on a single pipeline, this way of texturing, and pixel calculation show certain advantages. However, the actual ability to work is not a simple relation of multiplication and complex relationship is difficult to judge. ATI X1600 RV530 GPU latest, only four rendering pipelines, each pipeline has 12 pixel fill unit.

now for a pipelined defined "Pixel Shader (pixel shader) + function, the TMU (texture units) + ROP (rasterization engine, ATI called Render Back End). Simply put, Pixel Shader completed pixels , TMU responsible for texture rendering ROP is responsible for the final output of the pixel, therefore, a complete traditional pipelined means a Pixel Shader operation be completed in one clock cycle, the output of a texture and a pixel. GeForce 6600LE , a traditional four the pipelined architecture graphics cards (4X1) a completed 4 Pixel Shader operations per clock cycle, the output of four textures and four pixels. pipelining = Pixel Shader + TMU + ROP, the concept has always been GPU manufacturers advocacy . However, with the development of technology, 3D games started orientation this balance naturally broken.

the arithmetic instructions continued to increase to the fact that the status quo is based on the pixel shader program, ATI began to spare no effort to commitment to increase the number of pixel pipelines. X1600 for example, it has four true sense of the pipeline, but Pixel Shader, TMU and ROP 3:1:1 relationship. Specifically, X1600 12 Pixel Shader , TMU and ROP was only four, so this GPU core in a single cycle can be 12 Pixel Shader operator, output 4 textures and four pixels other words, X1600, Pixel Shader computing equivalent to 16 pipeline graphics, but the texture fill rate and pixel fill rate is equivalent to 4 pipelined graphics It stands to reason, X1600 technology is advanced, but a technical advantage as the weights of the extraordinary level of competition, which is obviously unreasonable the face of true 8-pipeline GeForce 6600GT and GeForce 7300GT, X1600 obviously no chance of winning. Fortunately, ATI recently released X1650XT, really have 12 pipelines and 36 pixel unit, but it seems that everything is already late. AMD ATI has acquired.

rely on NV40 NVIDIA punched a turnaround in mid-2005, the new G70 graphics chip successfully listed. radical NV40 to NV40, G70-based, in terms of performance and power consumption, cost and several other areas achieved a very good balance.

R520

eager to regain a city of ATI was continuous case of Re-tape the power consumption and other unfavorable factors, the legendary R520 repeatedly bounced after four months of suffering the G70 VS nothing of suffering, R520 was finally released in October. the

R420, like only 16 rendering pipeline, assigned processor with extremely threads, R520 up to handle 512 threads, thread management mechanisms so that each greatly enhance the efficiency of the rendering pipeline; 8 introduces SM3.0 vertex shaders, dynamic flow control instruction support, bypassing the SM3.0 provisions of VTF using R2VB way; uses a 256-bit ring bus increased memory latency, flexible scheduling data; support FP32 HDR + AA; advanced Avivo technology makes the ATI video quality on a new level.

R580

compared with the G70, the R520 technology advanced pixel shader unit is too small is flawed, launched soon after NVIDIA 7800GTX 512MB by virtue of the high-frequency and high-capacity memory easily exceeds R520 R520 disadvantage of ATI quickly responded with R520 launch very different architecture R580.

ATI think the next game will be on the higher requirements of Shader pixel shaders and TMU ratio should be larger, so R580 48 3D +1 D pixel shader unit, was used and R520 same 16TMU. this strange 3:1 architecture proved in Need for Speed ​​10 and obtained a more excellent performance than the traditional 1:1 architecture Elder Scrolls 4 PS resources tight game state-of-the-art soft shadow filtering technology Fetch4 R580 shadow processing more efficient.

R580 comparable performance opponent's flagship product in the old game, especially high-resolution AA/AF fully open Shader heavy pressure in the new game mode can even exceed more than 50% of the performance of rival products. compared to the R520, R580 game performance is greatly enhanced, and became the king of the game.

However, the R580 series and its predecessors R420 has a common weakness that low-end products in the same poor performance. Following the NV43 swept in low-end, successor G73 plays the same role; at the same time high of 20 pipeline 7900GS chip yield into the history of the lowest cost in the high-end chip, continuous price cuts to powerful high-end products in the ATI pressure. contrast, ATI X1600 series R580 1/4 of a smaller version of the storming of the low-end market, 12PS 4TMU this outstanding high-end chip 3: 1 architecture design somewhat acclimatized in the resolution of common games, too little of the TMU resources become performance bottlenecks ATI in the low-end of the competition once again defeat

R580 +

on when AMD is about to complete the acquisition of the R580 + listed The listing brings a new series of product lines to high-end playing mainly X1950XTX, basically can be seen as the GDDR4 version of the X1900XTX the GDDR4 applications greatly enhance the transmission bandwidth, to some extent also improve Performance. RV570 that X1950Pro/GT 36 PS unit, 12TMU, 12ROPs, the same 3:1 architecture, in many respects, than rival products, to become the most powerful high-end market competitors. RV560 that X1650GT is can be seen as R580 1/2, 24PSU, 8TMU the, 8ROPs test scores more than G73, has become one of the most cost-effective low-end graphics chip.

unfortunately X1950XTX full range of products out too late, despite good positioning and strong performance, it was too late, ATI has been the end of their mission on behalf of the completion of the rest will be AMD

RV630

2007 mid-April, NVIDIA officially announced its three end DirectX 10 graphics card to the outside world - the GeFroce 8500GT, 8600GT and 8600GTS. means, nVIDIA occupy the high end DX9 graphics card market, is also moving to the mainstream DX10 graphics card market in mid-May 2007, ATi launched its first flagship DX10 graphics card - Radeon HD 2900XT, but do not have the backbone, AMD-ATi not always in the mainstream market foothold So AMD-ATi June 12, 2007, the ED also announce its low-end to mid-range mainstream DX10 graphics card --RV630/610 series.

off than the first batch nVIDIA end DX10 graphics card slow a beat, but the AMD-ATi RV630 seems stamina sharp, thanks to the more advanced 65nm core process development vendors to ATi Radeon HD 2600 series good cost control, in accordance with the different positioning of the inexpensive Radeon HD 2600Pro overclocked version of the Radeon HD 2600Pro, nonpublic version 2600XT ..... variety of different specifications the RV630 graphics landed market has become ATi DX10 market in mid burst on the scene

on a large team.

R600 edge finally, in 2006, AMD-ATI released their first support DX10, SM4.0 graphics card - ATi Radeon HD 2900XT. HD2900XT based on the R600 core, 80nm manufacturing craft, she has a luxurious variety of hardware configurations, which can be seen after the acquisition of AMD-ATi devoted a great deal of hard work she has 720 million transistors, 320 superscalar stream processors, 512-bit display save bits wide, the largest more than 128.5GB/s of memory bandwidth, full support for DX10, SM4.0, AvivoHD video hardware accelerator components. numerous gaming performance tests, HD2900 are lost to competitors the nVIDIA based on the G80 core 8800 series.

mainly due to the following points:

1 ATi DX10 support on selected superscalar stream processors stream processor (SP), which helps to deal with a large number of parallel data, but superscalar SP 5 group will play all the utility. nVIDIA selected vector SP can play all utility so that compared to the number of truly comparable SP HD2900XT 64 (X5) a; 8800Ultra/GTX128, 8800GTS (640MB) 118, 8800GTS (320MB) 96 Ge this one than to know HD2900 even if there is fabulous as it is impossible to defeat 8800 series. entire HD2000 series are 24 (X5) a 2600, 2400 8 (X5) a.

2 ATi does not support core Frequency asynchronous, that is the core frequency must be followed by the SP frequency nVIDIA SP frequency k can easily go beyond 1GHz, we can see that ATi HD2000 series can not compete at the same level.

3, ATi is not properly play the core Characteristics. R600 core ring bus architecture, the data read speed, hit rate data characteristics determines the the HD2900 needs of high-speed memory, rather than a high-bandwidth memory. the HD2900 use 512-bitGDDR3 memory, like one eye is not superior power in a very wide, but resistance is great on the floor looking for something like 1GB version HD2900XT GDDR4 memory itself, the delay time is too long to be a drag.

4 HD2000 series of algorithms is not enough mature, BUG lot of open AA (anti-aliasing) performance significantly decreased.

Nevertheless HD2000 series does not fall too G8000 Series, HD2000 series high efficiency can be seen. gratifying

HD2000 the fringe of AvivoHD hard decoding technology effect than the G8000 series. contrast, HD2000, there is a certain competitiveness.

In order to enhance the future Radeon HD 2600/2400 Series Features advantage, ATi in RV630/610 are joined G84/G86 more powerful than Avivo HD video playback technology, the UVD Universal video decoding engine, while achieving level decoding of H.264 and VC-1 encoded video hardware graphics acceleration cards continue to enrich in CPU computing power dependence video playback capabilities to further the process of the release of high-definition video playback. ATi powerful HD playback ability ATi Radeon HD 2400 series of major stakes in the entry market.

R600 maturity - HD3800 Series

2007 November 16, AMD-ATi HD3800 series graphics cards she is the industry's first 55nm process graphics, the industry's first graphics card support DirectX10.1, Shader Model4.1 her based on RV670 core retained the R600 core and most of the characteristics of the same 320 superscalar SP reduced to 666 million, the number of transistors she introduced the new technology of ATi, such as PowerPlay energy-saving technology, AvivoHD second-generation hardware decoding technology. AMD-ATi This does not ATi's Radeon HD 3800 series, HD3800 who use 512-bit memory, but only with with 256-bitGDDR3 memory, visible AMD-ATi HD2900 body memory strategies error has been found.

advance support for Microsoft upcoming release of DirectX 10.1, but also its technical highlights one of ATI's RV670 and original nVIDIA G80 at this point is very similar to the earlier version of DirectX support. while, DX10.1 maintaining a 10 versions of the original overall structure and programming system, vertex, geometry and pixel shader instruction set will be updated to support for Shader Model 4.1, DirectX 10.1 also provides a number of enhancements, such as improved HDR lighting effects, anti-aliasing improvements, more accurate pipeline, global illumination effects such as DirectX 10.1 is an enhanced version of the DirectX 10 version, sold in the plan with Windows Vista SP1.

for ATi is also based on the RV670 core Radeon HD 3850, HD3870 is undoubtedly the weight of weight, RV670 listed so far, a variety of non-public version of the design DDR4 version of the Radeon HD 3850 also appears above in the market, their prices are maintained at less than $ 1,500, the HD 3850 listed for nVIDIA, after a long period of time not really come up with a product to combat ATi from 1500 yuan on the market impact of this embarrassing situation, HD3800 series nVIDIA impact of a blow so the nVIDIA hastily G94 9800GT 9600GT in order to cope with the HD3800 series but also the first Last positioning high-end graphics card lowest price times at around 2,000 yuan,

just listed the 9800GT, 9600GT is the historic fast Diving break the thousand mark the pinnacle of

R600 - HD4800 series

2008 June, AMD-ATi experience more than a year of hardships, finally broke! based on the RV770 core Radeon HD4800 series is finally released! the preemptive listing the HD4850 ​​has 1T (10 ^ 12) Flops floating point computing power, easily beat the high-end market GeForce9800GTX, even if nVIDIA released high frequency 9800GTX + also avail. RV770 RV670 comprehensive upgrade version, 320 superscalar SP mature manufacturing processes, increased to 800 superscalar SP, grating and texture unit increase, AMD-ATi HD4800 series who use AA optimized algorithm able to win the game open AA G9800 series graphics cards. listed later AMD-ATi HD4870 body using the correct memory strategy, introduced with 256-bitGDDR5 memory, return to the traditional display memory controller design, and the use of the frequency reached 3.6GHz! HD4870 performance on a layer, with 1.2TFlops the industry's first floating-point computing power, game performance even with monster GTX260 strong words.

ATI HD 4850 graphics

Radeon HD 4800 series outstanding performance, just a listing of 1499 yuan HD4850 ​​pricing day forcing nVIDIA GeForce9800GTX dropped from 2,999 yuan to 1,999 yuan a few days after the approaching 1399 yuan. HD4800 series of cost-effective, we can see its influence outstanding by the graphics card hobby sought after hot in all countries, and some countries even once out of stock 12 January

8,, AMD-ATi released HD4870X2 and 4850X2. 4870X2 face flagship market, against the most senior monster graphics the GTX280. HD4850X2 is to encourage the main push of non-public version of the product, the cost savings to reduce prices to occupy the high-end market day, AMD-ATi released drivers CatalystControlCenter8.8, a single card will be dual-core and dual-card interconnect solutions

没有评论:

发表评论