GPU Evolution: ATI’s Strategy

ATI’s habitual way of behaving in the desktop graphics market looks like that: they release a new graphics processor and graphics cards based on it. Then, some time later, a new version of the chip appears, usually with the same architecture, but capable of working at a higher frequency thanks to the improved tech process. They also launch a junior model, specifically for lower market tiers. Thus, ATI’s product range is constantly expanding, without any big effort on the company’s side.

You want examples? Just recall such successful modifications of the RADEON 9800 PRO and RADEON 9600 PRO as the RADEON 9800 XT and 9600 XT, respectively. The graphics processors called R360 and RV360 didn’t differ in anything from the R350 and RV350, respectively, except that they could work at higher frequencies, so ATI Technologies announced two new products with a higher performance.

Sometimes ATI just replaces one GPU with two different ones, intended for two different market sectors. Recalling the RADEON 9700/9500, the ATI R300 chip was first employed to make top-end (RADEON 9700 PRO) as well as mainstream (RADEON 9500/9500 PRO) graphics processors. In the latter case, the graphics cards were equipped with a 128-bit memory bus and had some of their pixel pipelines disabled (in the RADEON 9500). Of course, it’s not profitable to make mainstream graphics cards using a complex and expensive graphics chip, but ATI had no choice actually – the company needed mass and high-performance solutions. Moreover, some dies were defective anyway, so why not use them? But the tech process getting perfected, and the number of the defective dies diminishing, the company at last decided to split top-end and mainstream product lines by releasing two independent graphics processors: RADEON 9800 PRO and RADEON 9600 PRO.

Thus, ATI trumpets the release of a new graphics architecture with a release of new high-end GPUs, and while these high-performance chips are attacking the top of the market, a special team of ATI engineers are tailoring the available high-end hardware to create something that suits a typical PC user. This policy has its downside, though. On the one hand, ATI has much time to adapt the graphics architecture for the needs of each particular market sector (high-end, mainstream, entry-level), getting a relatively inexpensive chip with an appropriate performance for each market segment. On the other hand, the adaptation and the design of the mainstream and entry-level chips take some time and this doesn’t allow the users who spend $100 for their graphics cards to have the same capabilities as those who pay $400-500. But practice shows that ATI users can usually choose from a dense group of same-functionality products priced within a range of $150-500 after 4-6 months since the release of a new GPU architecture.