ATI Technologies: Gone But Not Forgotten

For 25 years, one title was synonymous with laptop graphics. By its trials and tribulations, it fought in opposition to different long-standing corporations in addition to sizzling new upstarts. Its most profitable product model remains to be in use right this moment, alongside lots of the applied sciences it pioneered through the years.

We’re, in fact, referring to ATI Applied sciences. Here is the fascinating story of its start, rise to prominence, and the last word conclusion of this Canadian graphics big.

A brand new star is born

As with many of the late, nice know-how corporations we have lined in our “Gone However Not Forgotten” collection, this story begins within the mid-Nineteen Eighties. Lee Lau, who had emigrated to Canada as a toddler, graduated with a Grasp’s diploma in digital engineering from the College of Toronto (UoT).

After initially working for a sub-division of Motorola within the early Nineteen Eighties, he established a small agency known as Comway that designed primary growth playing cards and graphics adapters for IBM PCs, with the manufacturing course of executed in Hong Kong.

This enterprise introduced him into contact with Benny Lau, a circuit design engineer and fellow UoT graduate, and Kwok Yen Ho, who had amassed a few years of expertise within the electronics business. At Lau’s urging, they joined him in reworking Comway into a brand new enterprise. With roughly $230,000 in financial institution loans and investments (some from Lee’s uncle, Francis Lau), the workforce formally based Array Expertise Integrated in August 1985.

ATI’s first product. Credit score: VGA Museum

Its mission was to solely design video acceleration chipsets primarily based completely on gate arrays (therefore the corporate title) and promote them to OEMs to be used in IBM PCs, in clones, and their very own growth playing cards. Resulting from authorized stress from a equally named enterprise within the USA, it adopted the title ATI Applied sciences inside just a few months of the corporate’s inception.

In lower than 12 months, it had its first product available on the market. The Graphics Resolution was a $300 card that would match into any free 8-bit ISA slot, and supported a number of graphics modes – it might simply change between MDA, CGA, Hercules, and Plantronics show outputs with the accompanying software program.

For a small further value, prospects might additionally purchase variations that included parallel and/or serial ports for connecting different peripherals to the PC. The one factor it did not supply was IBM’s EGA mode, however given the value, nobody was actually complaining.

Credit score: VGA Museum

Being a newcomer within the business, gross sales had been maybe modest, however with steady funding and engineering prowess, the group continued to reinforce their chips. Because the Nineteen Eighties neared its finish, ATI proved it was not only a fleeting phenomenon – the $400 EGA Surprise (1987) and $500 VGA Surprise (1988, pictured above) showcased substantial enhancements over their predecessors and had been competitively priced.

Through the years, a number of variations of those merchandise had been launched. Some had been ultra-affordable choices with minimal options, whereas others featured devoted mouse ports or sound chips on the circuit board. Nevertheless, this was a extremely aggressive market, and ATI wasn’t the one firm producing such playing cards.

What was wanted was one thing particular, one thing that may make them stand out from the gang.

Mach enters the stage

Graphics playing cards within the 80s had been vastly totally different from what we all know right this moment, regardless of any superficial similarities. Their major perform was to transform the show info generated by the pc into {an electrical} sign for the monitor – nothing extra. All the computations concerned in truly producing the displayed graphics and colours had been completely dealt with by the CPU.

Nevertheless, the time was ripe for a devoted co-processor to take over a few of these duties. One of many first entries available on the market got here from IBM, within the type of its 8514 show adapter. Whereas more cost effective than workstation graphics playing cards, it was nonetheless fairly expensive at over $1,200, naturally prompting a slew of corporations to provide their very own clones and variations.

Credit score: Retronn.de

Within the case of ATI, this took the form of the ATI38800 chip – higher referred to as the Mach8. Implementing this processor in a PC led to sure 2D graphics operations (e.g., line drawing, space filling, bit blits) being delegated from the CPU. One main downside of the design, nonetheless, was its lack of VGA capabilities. This compelled the necessity for a separate card to deal with this process or contain an extra chipset on the cardboard.

The ATI Graphics Extremely and Graphics Vantage, launched in 1991, had been two such fashions that took the latter route. Nevertheless, they had been considerably costly for what they provided. As an example, the top-tier Extremely, geared up with 1MB of dual-ported DRAM (marketed beneath the moniker of VRAM), value $900. Costs shortly plummeted as soon as ATI unveiled its subsequent graphics processor.

The ATI Graphics Extremely and Graphics Vantage, launched in 1991, had been two such fashions that took the latter route, however they had been fairly expensive for what they provided. For instance, the top-end Extremely, loaded with 1MB of dual-ported DRAM (marketed beneath the moniker of VRAM), value $900, although costs shortly plummeted as soon as ATI unveiled its subsequent graphics processor.

Packed stuffed with enhancements, the ATI68800 (Mach32) reached cabinets in late 1992. With an built-in VGA controller, 64-bit inner processing, as much as 24-bit colour help, and {hardware} acceleration for the cursor, it was a big step ahead for the corporate.

Credit score: VGA Museum

The Graphics Extremely Professional (above) sported the brand new chip, together with 2 MB of VRAM and an asking value of $800 – it was genuinely highly effective, particularly with the optionally available linear aperture characteristic enabled, which allowed purposes to write down knowledge instantly into the video reminiscence. Nevertheless, it was not with out points, very similar to its opponents. ATI provided a extra economical Plus model with 1 MB of DRAM and no reminiscence aperture help. This mannequin was much less problematic and extra standard on account of its $500 price ticket.

Over the subsequent two years, ATI’s gross sales improved amidst fierce competitors from the likes of Cirrus Logic, S3, Matrox, Tseng Laboratories, and numerous different companies. In 1992, a subsidiary department was opened in Munich, Germany, and a 12 months later, the corporate went public with shares offered on the Toronto and Nasdaq Inventory Exchanges.

Nevertheless, with quite a few corporations vying in the identical enviornment, sustaining a constant lead over everybody else was extraordinarily difficult. ATI quickly confronted its first setback, posting a internet lack of just below $2 million.

It did not assist that different corporations had more cost effective chips. Round this time, Cirrus Logic had bought a tiny engineering agency, known as AcuMOS, which had developed a single-die resolution, containing the VGA controller, RAMDAC, and clock generator – ATI’s fashions all used separate chips for every function, which meant the playing cards value extra to fabricate.

Credit score: Retronn.de

Nearly two years would go earlier than it had an acceptable response, within the type of the Mach64. The primary variants of this chip nonetheless used an exterior RAMDAC and clock, however these had been swiftly changed by totally built-in fashions.

This new graphics adapter went on to encourage a mess of product strains, significantly within the OEM sector. Xpression, WinTurbo, Professional Turbo, and Charger all turned acquainted names in each family and workplace PCs.

Regardless of the delay in launching the Mach64, ATI wasn’t idle in different areas. The founders had been constantly observing the competitors and foresaw a shift within the graphics market. They bought a Boston-based graphics workforce from Kubota Company and tasked them with designing the subsequent technology of processors to grab the chance as soon as the change started, whereas the remainder of the engineers continued to refine their present chips.

Rage contained in the machine

Two years after launching its Mach64-powered playing cards, ATI introduced a brand new processor into the limelight in April 1996. Codenamed Mach64 GT, however marketed as 3D Rage, this was the corporate’s first processor providing each 2D and 3D acceleration.

On paper, the 3D Xpression playing cards that sported the chip appeared to be actual winners – help for Gouraud shading, perspective-correct texture mapping (PCTM), and bilinear texture filtering had been all current. Operating at 44 MHz, ATI claimed the 3D Rage might obtain a peak pixel fill fee of 44 Mpixels/s, although these efficiency claims had a number of caveats.

The 3D Rage was developed with Microsoft’s new Direct3D in thoughts. Nevertheless, this API was not solely delayed but in addition underwent fixed specification modifications earlier than its eventual launch in June 1996. In consequence, ATI needed to develop its personal API for the 3D Rage, and solely a handful of builders adopted it.

ATI’s processor did not help z-buffering, that means any recreation utilizing it would not run, and the alleged pixel fill fee was solely achievable with flat-shaded polygons or these coloured with non-perspective-correct texture mapping. With bilinear filtering or PCTM utilized, the fill fee would halve, on the very least.

Credit score: VGA Museum

To make issues worse, the mere 2 MB of EDO DRAM on the early 3D Xpression fashions was sluggish and restricted 16-bit colour rendering to a most decision of 640 x 480. And to cap all of it off, even in video games it might run, the 3D efficiency of the Rage processor was, at greatest, underwhelming.

Luckily, the product did have some positives – it was moderately priced at round $220, and its 2D and MPEG-1 acceleration had been commendable for the value. ATI’s OEM contracts ensured that many PCs had been offered with a 3D Xpression card inside, which helped keep a gradual income stream.

An up to date model of the Mach64 GT (aka the 3D Rage II) appeared later the identical 12 months. This iteration supported z-buffer and palletized textures, together with a variety of APIs and drivers for a number of working methods. Efficiency improved, due to increased clock speeds and a bigger texture cache, however was nonetheless underwhelming, particularly in comparison with the choices of latest opponents.

Launched in October 1996, 3dfx Voodoo Graphics was 3D-only, however its rendering chops made the Rage II look decidedly second-class. None of ATI’s merchandise got here near matching it.

ATI was extra invested within the broader PC market. It developed a devoted TV encoding chip, the ImpacTV, and together with the addition of MPEG-2 acceleration within the 3D Rage II chip, customers might buy an all-in-one media card for a comparatively modest sum.

This development continued with the subsequent replace in 1997, the 3D Rage Professional, regardless of its enhanced rendering capabilities. Nvidia’s new Riva 128, when paired with a reliable CPU, considerably outperformed ATI’s providing.

Nevertheless, playing cards such because the Xpert@Play appealed to budget-conscious consumers with their complete characteristic units. The All-In-Surprise mannequin exemplified this (proven beneath). It was a regular Rage Professional graphics card geared up with a devoted TV tuner and video seize {hardware}.

Credit score: VGA Museum

As the event of graphics chips accelerated, smaller companies struggled to compete, and ATI acquired a former competitor, Tseng Laboratories, on the finish of the 12 months. The inflow of latest engineers and experience resulted in a considerable replace to the Rage structure in 1998 – the Rage 128. However by this time, 3dfx Interactive, Matrox, and Nvidia all had spectacular chip designs and a variety of merchandise available on the market. ATI’s designs simply could not compete on the prime finish.

Nonetheless, ATI’s merchandise discovered success in different markets, significantly within the laptop computer and embedded sectors, securing a stronger income stream than every other graphics agency. Nevertheless, these markets additionally had slim revenue margins.

The late Nineties noticed a number of modifications throughout the firm – Lee Lau and Benny Lau departed for brand spanking new ventures, whereas further graphics corporations (Chromatic Analysis and part of Real3D) had been acquired. Regardless of these modifications, chip improvement continued unabated, leading to what would finally be the penultimate model of the 3D Rage featured in a brand new spherical of fashions.

Credit score: Retronn.de

The Rage 128 Professional chip powered Rage Fury playing cards for the gaming sector and Xpert 2000 boards for workplace machines. ATI even mounted two processors on one card, the Rage Fury MAXX (above), to compete in opposition to the perfect from 3dfx and Nvidia.

On paper, the chips on this specific mannequin had the potential to supply virtually double the efficiency of the usual Rage Fury. Sadly, it simply wasn’t nearly as good as Nvidia’s new GeForce 256 and it was costlier; it additionally lacked help for {hardware} acceleration vertex remodel and lighting calculations (TnL), a brand new characteristic of Direct3D 7 at the moment.

It was clear that if ATI was going to be seen as a market chief in gaming, it wanted to create one thing particular – once more.

A brand new millennium, a brand new buy

With the model title Rage related to cheap but competent playing cards, ATI started the brand new millennium with important modifications. They simplified the outdated structure codename – Rage 6c turned R100 – and launched a brand new product line: the Radeon.

Nevertheless, the modifications had been extra than simply skin-deep. They overhauled the graphics processing unit (GPU, because it was more and more being known as), turning it into one of the superior consumer-grade rendering chips of the time.

One vertex pipeline, totally able to {hardware} Rework and Lighting (TnL), fed into a posh triangle setup engine. This engine might kind and cull polygons primarily based on their visibility in a scene. In flip, this fed into two pixel pipelines, every housing three texture mapping items.

The unique Radeon graphics card. Credit score: {Hardware} Museum

Sadly, its superior options proved a double-edged sword. Probably the most generally used rendering API for video games at the moment, Direct3D 7, did not help all its capabilities, and neither did its successor. Solely by way of OpenGL and ATI’s extensions might its full capabilities be totally realized.

The primary Radeon graphics card hit the market on April 1st, 2000. Regardless of potential jokes in regards to the date, the product was no laughing matter. With 32 MB of DDR-SDRAM, clock speeds of 166 MHz, and a retail value of round $280, it competently rivaled the Voodoo5 and GeForce2 collection from 3dfx and Nvidia, respectively.

This was actually true when taking part in video games in 32-bit colour and at excessive decision, however at decrease settings and 16-bit colour, the competitors was quite a bit stronger. Preliminary drivers had been considerably buggy and OpenGL efficiency wasn’t nice, both, though each of those facets improved in the end.

Over the subsequent 12 months, ATI launched a number of variants of the R100 Radeon. One used SDRAM as a substitute of DDR-SDRAM (cheaper however much less highly effective), a 64 MB DDR-SDRAM model, and a funds RV100-powered Radeon VE. The RV100 chip was basically a faulty R100, with many components disabled to stop operational bugs. It was priced competitively in opposition to Nvidia’s GeForce2 MX card, however like its mum or dad processor, it could not fairly match up.

Nvidia’s GeForce2 lineup held the efficiency crown. Supply: Wikimedia

Nevertheless, ATI nonetheless dominated the OEM market, and its coffers had been full. Earlier than the Radeon hit the market, ATI set out as soon as once more to amass promising prospects. They bought the lately shaped graphics chip agency ArtX for a big sum of $400 million in inventory choices. This might sound uncommon for such a younger enterprise, however ArtX stood out.

Composed of engineers initially from Silicon Graphics, the group was behind the event of the Actuality Sign Processor within the Nintendo 64 and was already contracted to design a brand new chip for the console’s successor. ArtX even had expertise creating an built-in GPU, which was licensed by the ALi Company for its funds Pentium motherboard chipset (the Aladdin 7).

By this acquisition, ATI did not simply safe recent engineering expertise but in addition secured a contract for producing the graphics chip for Nintendo’s GameCube. Primarily based in Silicon Valley, the workforce was tasked with aiding the event of a brand new, extremely programmable graphics processor, though it might take just a few years earlier than it was prepared.

The R200-powered Radeon 8500. Credit score: {Hardware} Museum

ATI continued its sample of making technically spectacular but underperforming merchandise in 2001 with the discharge of the R200 chip and the subsequent technology of Radeon playing cards, now following a brand new 8000-series naming scheme.

ATI addressed all of the shortcomings of the R100, making it totally compliant with Direct3D 8, and added much more pipelines and options. On paper, the brand new structure provided greater than every other product available on the market. It was the one product that supported pixel shader v1.4, elevating expectations.

However as soon as once more, exterior of artificial benchmarking and theoretical peak figures, ATI’s greatest consumer-grade graphics card merely wasn’t as quick as Nvidia’s. This actuality started to indicate between 1999 and 2001. Regardless of accumulating $3.5 billion in gross sales, ATI recorded a internet lack of almost $17 million.

In distinction, Nvidia, its important competitor, earned lower than half of ATI’s income however made a internet revenue of $140 million, largely on account of promoting a big quantity of high-end graphics playing cards with a lot bigger revenue margins than primary OEM playing cards.

A part of ATI’s losses stemmed from its acquisitions – after ArtX, ATI bought a piece of Appian Graphics (particularly, the division creating multi-monitor output methods) and Diamond Multimedia’s FireGL model and workforce. The latter offered a foothold within the workstation market, as FireGL had constructed a stable status in its early years utilizing 3DLabs’ graphics chips.

However ATI could not merely spend its method to the highest. It wanted greater than only a wealth of high-revenue, low-margin contracts. It wanted a graphics card superior in each doable method to the whole lot else available on the market. Once more.

The killer card lastly arrives

All through the primary half of 2002, rumors and purported feedback from builders relating to ATI’s upcoming R300 graphics chip started to flow into on the web.

The design of this next-generation GPU was closely influenced by the previous ArtX workforce. Among the figures being touted – like double the variety of transistors in comparison with the R200, 20% increased clock speeds, 128-bit floating level pixel pipelines, and anti-aliasing with almost no efficiency hit – appeared virtually too good to be true.

Naturally, as folks usually do, most dismissed these claims as mere hyperbole. In any case, ATI had garnered a status for over-promising and under-delivering.

One of many all-time greats in GPU historical past. Credit score: PixelPipes

However when the $400 ATI Radeon 9700 Professional appeared in August, sporting the brand new R300 at its coronary heart, the truth of the product was an enormous shock. Whereas not all rumors had been completely correct, all of them underestimated the cardboard’s energy. In each evaluate, ATI’s new mannequin fully outclassed its opponents; neither the Matrox Parhelia-512 nor the Nvidia GeForce4 Ti 4600 might compete.

Each side of the processor’s design had been meticulously fine-tuned. This ranged from the extremely environment friendly crossbar and 256-bit reminiscence bus to the primary use of flip-chip packaging for a graphics chip to facilitate increased clocks.

With twice the variety of pipelines because the 8500 and full compliance with Direct3D 9 (nonetheless just a few months from launch), the R300 was an engineering marvel. Even the drivers, a long-standing weak level for ATI, had been steady and wealthy in options.

The Radeon 9700 Professional and the slower, third-party-only Radeon 9700 had been instantaneous hits, bolstering revenues in 2003. For the primary time in three years, the online revenue was additionally optimistic. Nvidia’s NV30-based GeForce FX 5800 collection, launched within the spring, additional aided ATI – it was late to market, power-hungry, sizzling, and marred by important obfuscation by Nvidia.

The NV30, just like the R300, purportedly had eight pixel pipelines. Nevertheless, whereas ATI’s processed colours at a 24-bit floating-point stage (FP24, the minimal required for Direct3D 9), Nvidia’s had been supposedly a full 32-bit. Besides, that wasn’t completely true.

Not one of many all-time greats. Credit score: {Hardware} Museum

The brand new GPU nonetheless had 4 pipelines, every with two texture items, so how might it declare eight? That determine accounted for particular operations, primarily associated to z-buffer and stencil buffer learn/writes.

In essence, every ROP might deal with two depth worth calculations per clock – a characteristic that may finally change into normal for all GPUs. The FP32 declare was an identical story; for colour operations, the drivers would usually power FP16 as a substitute.

In the fitting circumstances, the GeForce FX 5800 Extremely outperformed the Radeon 9700 Professional, however more often than not, the ATI card proved superior – particularly when utilizing anti-aliasing (AA) and anisotropic texture filtering (AF).

The designers persistently refined the R300, and every successive model (R420 in 2004, R520 in 2005) went head-to-head with the most effective from Nvidia – stronger in Direct3D video games, a lot stronger with AA+AF, however slower in OpenGL titles, particularly these utilizing id Software program’s engines.

Wholesome revenues and income characterised this era, and the balanced competitors meant that customers benefited from wonderful merchandise throughout all funds sectors. In 2005, ATI acquired Terayon Communication Programs to bolster its already dominant place in onboard TV methods and, in 2006, bought BitBoys Oy to strengthen its foothold within the handheld units business.

An early style of the way forward for GPUs. Credit score: Oleg Kashirin

ATI’s R300 design caught Microsoft’s consideration, main the tech big to contract ATI to develop a graphics processor for its subsequent Xbox console, the 360. When it launched in November 2005, the path of future GPUs and graphics APIs turned clear – a completely unified shader structure the place discrete vertex and pixel pipelines would merge right into a single array of processing items.

Whereas ATI was now one step forward of its opponents on this space, it nonetheless wanted to maintain its development. The answer to this thorny downside got here as an almighty shock.

All change on the prime

In 2005, the co-founder, long-standing Chairman, and CEO, Kwok Yuen Ho, retired from ATI. He, his spouse, and a number of other others had been beneath investigation for insider buying and selling of firm shares for 2 years, courting again to 2000. Though the couple was exonerated from all costs, ATI itself was not, paying virtually $1 million in costs for deceptive the regulators.

Ho had already stepped down as CEO by then. David Orton, former CEO of ArtX, took over in 2004. James Fleck, an already rich and profitable businessman, changed Ho as Chairman. For the primary time in twenty years, ATI had a management construction with none connections to its origins.

In 2005, the corporate posted its highest-ever revenues – barely over $2.2 billion. Whereas its internet revenue was a modest $17 million, ATI’s sole competitor out there, Nvidia, was not faring a lot better, regardless of having a bigger market share and larger income.

With this in thoughts, the brand new management was assured in realizing its objectives of boosting income development by creating its weaker sectors and consolidating its strengths.

In July 2006, virtually fully out of the blue, AMD introduced its intention to amass ATI. The alternate was for $4.2 billion in money (60% of which might come from a financial institution mortgage) and $1.2 billion of its shares. AMD had initially approached Nvidia however rejected the proposed phrases of the deal.

The acquisition was a colossal gamble for AMD, however ATI was elated, and for a great cause. This was a chance to push Nvidia out of the motherboard chipset market. ATI might probably use AMD’s personal foundries to fabricate a few of its GPUs, and its designs could possibly be built-in into CPUs.

AMD was evidently prepared to spend a lot of cash and had the monetary backing to take action. Better of all, the deal would additionally imply that ATI’s title would stay in use.

The acquisition was finalized by October 2006, and ATI formally turned a part of AMD’s Graphics Product Group. Naturally, there was some reshuffling within the administration construction, nevertheless it was in any other case enterprise as standard. The primary consequence was the discharge of the R600 GPU in Could 2007 – ATI’s first unified shader structure, referred to as TeraScale, for the x86 market.

The brand new design was clearly primarily based on the work achieved for the Xbox 360’s graphics chip, however the engineers had scaled the whole lot significantly increased. As a substitute of getting devoted pipelines for vertex and pixel shaders, the GPU was made up of 4 banks of sixteen Stream Processor Items (every containing 5 Arithmetic Logic Items, or ALUs) that would deal with any shader.

Boasting 710 million transistors and measuring 420 mm2, the R600 was twice the dimensions of the R520. The primary card to make use of the GPU, the Radeon HD 2900 XT, additionally had spectacular metrics – 215 W energy consumption, a dual-slot cooler, and 512 MB of quick GDDR3 on a 512-bit reminiscence bus.

Nvidia had crushed ATI to the market by launching a unified structure (referred to as Tesla) a 12 months earlier. Its G80 processor, housed within the GeForce 8800 Extremely, was launched in Could 2006. Nvidia then adopted up with scaled-down GTX and GTS fashions in November. When matched in precise video games, ATI’s greatest was slower than the Extremely and GTX, however higher than the GTS.

The Radeon HD 2900 XT’s saving grace was its value. At $399, it was $200 cheaper than the GeForce 8800 GTX’s MSRP (although this hole narrowed to about $100 by the point the R520 card was launched) and a staggering $420 lower than the Extremely. For the cash, the efficiency was excellent.

Why wasn’t it higher? Among the efficiency deficit, in comparison with Nvidia’s fashions, could possibly be attributed to the truth that the R520 had fewer Texture Mapping Items (TMUs) and Render Output Items (ROPs) than the G80. Video games on the time had been nonetheless closely depending on texturing and fill fee (sure, even Crysis).

TeraScale’s shader processing fee was additionally extremely depending on the compiler’s means to correctly cut up up directions to reap the benefits of the compute construction. In brief, the design was progressive and forward-thinking (the usage of GPUs in compute servers was nonetheless in its infancy), nevertheless it was not excellent for the video games of that interval.

ATI had demonstrated that it was completely able to making a unified shader structure particularly for gaming with the Xenos chip within the Xbox 360. Due to this fact, the choice to focus extra on computation with a GPU will need to have been pushed by inner stress from senior administration.

After investing a lot in ATI, AMD was pinning excessive hopes on its Fusion challenge – the mixing of a CPU and GPU right into a single, cohesive chip that may enable each components to work on sure duties in parallel.

This path was additionally adopted with the 2008 iteration of TeraScale GPUs – the RV770. Within the top-end Radeon HD 4870 (above), the engineers elevated the shader unit depend by 2.5 occasions for simply 33% extra transistors.

The usage of quicker GDDR5 reminiscence allowed the reminiscence bus to be halved in measurement. Coupled with TSMC’s 55nm course of node, your entire die was virtually 40% smaller than the R600. All this resulted within the card being retailed at $300.

As soon as once more, it performed second fiddle to Nvidia’s newest GeForce playing cards in most video games. Nevertheless, whereas the GeForce GTX 280 was significantly quick, its launch MSRP was greater than double that of the Radeon HD 4870.

The period of ATI’s graphics playing cards being the most effective by way of absolute gaming efficiency was gone, however nobody was beating them when it got here to worth for cash. Not like the 3D Rage period when the merchandise had been low cost and sluggish, the final Radeons had been reasonably priced however nonetheless delivered passable efficiency.

Within the late summer time of 2009, a revised TeraScale structure launched a brand new spherical of GPUs. The household was known as Evergreen, whereas the precise mannequin used within the high-end Radeon HD 5870 was referred to as Cypress or RV870. Whatever the naming confusion, the engineers managed to double the variety of transistors, shader items, TMUs, and so on.

The gaming efficiency was wonderful, and the small enhance in MSRP and energy consumption was acceptable. Solely twin GPU playing cards, equivalent to Nvidia’s GeForce GTX 295 and ATI’s personal Radeon HD 4870 X2, had been quicker, although they had been naturally costlier. The HD 5870 solely confronted actual competitors six months later when Nvidia launched its new Fermi structure within the GeForce GTX 480.

Whereas not fairly a return to the 9700 Professional days, the TeraScale 2 chip confirmed that ATI’s designs had been to not be dismissed.

Goodbye to ATI?

In 2010, ATI had an intensive GPU portfolio – its discrete graphics playing cards spanned all value segments, and although its market share was round half of Nvidia’s, Radeon chips had been widespread in laptops from a wide range of distributors. Its skilled FireGL vary was thriving, and OEM and console GPU contracts had been strong.

The much less profitable facets had already been trimmed. Motherboard chipsets and the multimedia-focused All-in-Surprise vary had been discontinued just a few years earlier, and AMD had offered the Imageon chipsets, utilized in handheld units, to Qualcomm.

Regardless of all this, AMD’s administration believed that the ATI model did not have sufficient standing by itself, particularly compared to AMD’s title. Due to this fact, in August 2010, it was introduced that the three letters that had graced graphics playing cards for 25 years could be dropped from all future merchandise – ATI would stop to exist as a model.

The eternal house of ATI. Credit score: F Wong

Nevertheless, this was just for the model. The precise firm stayed energetic, albeit as a part of AMD. At its peak, ATI had over 3500 staff and a number of other world workplaces. At present, it trades beneath the title of ATI Applied sciences ULC, with the unique headquarters nonetheless in use, and its workforce totally built-in into related sectors in AMD’s organizational construction.

This group continues to contribute to the analysis and improvement of graphics know-how, as evidenced by patent listings and job vacancies. However why did not AMD simply fully take up ATI, like Nvidia did with 3dfx, as an illustration? The phrases of the 2006 merger meant that ATI would change into a subsidiary firm of AMD.

Whereas it is doable that ATI might have been finally shut down and totally absorbed, the extent of funding and near-endless court docket instances over patent violations that ATI has continued to take care of means it is not going wherever simply but, if ever.

Nevertheless, with a big quantity of GPU improvement work now carried out by AMD itself in numerous places worldwide, it is now not correct to label a graphics chip (whether or not in a Radeon card or the most recent console processor) as an ATI product.

The ATI model has now been absent for over a decade, however its legacy and the fond reminiscences of its outdated merchandise endure. The ATI Applied sciences department continues to contribute to the graphics subject almost 40 years after its inception and you might say it is nonetheless kicking out the products. Gone however not forgotten? No, not gone and never forgotten.