X-bit labs

The Strong Dedication to the Quality: Interview with John Byrne of AMD Graphics Product Group (page 2)

Category: Editorial

by Anton Shilov

[ 03/24/2007 | 04:18 PM ]

There are a lot of hype about products like ATI’s Radeon X1950 XTX or Nvidia’s GeForce 8800 GTX, but the most of the money (http://www.xbitlabs.com/news/video/display/20061229150853.html) come from performance mainstream offerings, the market segment, where ATI did not perform well all the time back in the days. The situation got much tougher when ATI’s high-end parts emerged months after Nvidia’s. Perhaps, the company knows how to improve such omissions?

X-bit labs: Products like the Radeon X700 and X1600 did not manage to achieve the success of Nvidia’s GeForce 6600, 6800 GS and 7600 GT. Are you working to avoid such situations in future?

John Byrne: Any comparative group test that includes the Radeon X1950 Pro will show you just how seriously we take the mid-range market. In a very competitive market, you will always have areas where you win easily and others where the competition put up more of a fight. In the end, it is this competition that drives the market forward and – ultimately – delivers a better experience for the consumer.

X-bit labs: The introduction of the code-named R600 will be the second time when ATI/AMD is several months late to market compared to the arch-rival Nvidia. Do you have a strategy to evade the same going forward?

John Byrne: Traditionally, at ATI, we have always had a strong dedication to quality – and this is an ethos that is shared with our colleagues throughout AMD. Quality means putting out the best solution possible – not rushing products to market. We are very confident in both our DirectX10 hardware and WHQL-certified drivers for XP and Vista.

X-bit labs: Do you plan to continue refreshing your product lineup every six months, or do you plan to change that in some ways?

John Byrne: That question focuses too narrowly on one segment of our roadmap. One of the things that most attracted AMD to ATI was the broad diversity of our product range.

In the graphics arena alone, we have not only the world’s most advanced graphics products for desktop systems, but also the world’s fastest consumer graphics for notebooks. At the same time, two out of three of the most advanced consoles also use ATI graphic products.

From there we can easily branch out to hand-held and embedded devices – as well as other offers for the workstation and server markets. With so many roadmaps and such a massive, on-going investment in R&D, it is easy to see why AMD’s graphics division is set to continue innovating at such a rapid pace.

Back in the days ATI introduced a brand-new graphics chip with brand-new functionality and feature-set for the high-end market and brought a little bit simplified version of the previous-gen product for the mainstream.

As a result, the company’s mainstream offerings were either a generation behind (in 2001 with Radeon 8500 and Radeon 7500, in 2002 with Radeon 9700 and 9100/9200 and in H1 2004 with Radeon X800 and X600/X300) or came a quarter or two later than the flagship offering (in 2003 with the Radeon 9700 and 9600 [the 9500-series was not really popular], in 2004 with the Radeon X800 and the Radeon X700). As a result of the Radeon X1800 (R520) delay, the company introduced a lineup of DirectX 9 shader model 3.0 supporting chips, but that was rather an exception, as low-cost Radeon X1900 (R580) derivative Radeon X1950 Pro/X1650 XT (RV570) came eight months later.

In fact, Nvidia has always had similar approach (with the exception of the GeForce FX, which was the case similar to the R520), but maybe introduction of GPUs with similar functionality, but with different performance and at different price-points makes sense too?

X-bit labs: Do you think it is strategically important to introduce a family of graphics processors for different markets, but which support similar functionality?

John Byrne: It is very important. As software development moves on to a new platform with the advent of Microsoft Windows Vista, so consumers will want to experience the new environment on PC and Apple, desktop and notebook. At the same time, developers using state-of-the-art ATI FireGL workstation cards will also be creating content targeted at the new API. Implementing a unified shader architecture across a range of devices, with rock solid drivers, will become increasingly important in the future.

Historically we have always focused on producing a high-end part first and then following up with a broad range which covers the remainder of the market.  The mid-range parts are typically much the most important to us in terms of revenue, but there’s no doubt that a good high end card whets people’s appetites.  In that sense the full range is important, but it’s OK to deliver the mid-range part a little later…

In terms of similar functionality there’s no doubt that it’s critical to us to provide the same core functionality in these later chips too.  As DirectX 10 games start to arrive in the middle of 2007 you’re going to start to see the convincing reasons to choose DX10 hardware.  At that point you can be confident that AMD’s range of hardware will support all these new games (and all your old favorites too).

X-bit labs: Do you think it makes sense to intro a DirectX 10 family of chips as soon as possible to the market, or it is better to start from the high-end while offering high-performance DX9 chips to the mainstream?

John Byrne: We’d love to be able to give everyone the choice – and for that reason we hope to be more aggressive in the way we plan to bring the full range of products to market than we have before now.  Against that is the fact that there are no DirectX 10 games ready to buy so far – so it’s very hard to see how to measure the real benefits of buying DX10 hardware.  When the games do arrive the hardware needs to be available, that’s one of the key reasons why it hasn’t been necessary for us to rush our products to market.  Instead we’ve had the relative luxury of being able to spend the time to make sure we have the right solutions.

X-bit labs: In fact, if you introduce a family of GPUs that perform relatively well in DirectX 10 environment, you would catalyze game developers to employ more DX10 effects, which would eventually catalyze people to upgrade to next-gen DX10 hardware quicker.  What do you think?

John Byrne: Making the big move to a completely new platform takes a lot of smaller steps. Content creation is intensely resource-hungry and time consuming. Then you have to allow for an installed user base to grow. At the same time, driver maturity is vital. The mass-market will be catalysed by the introduction of our complete range of DirectX 10 products – with fully certified WHQL drivers – which will help create a base of users that will prove attractive to developers. It is a virtuous circle that we all play a small part in creating/energising.

Games developers don’t really need us as a catalyst.  The transition to DirectX10 is a major jump – so it’s true that there are quite a few games developers who have decided to stand back.  But there are also many developers who have embraced the new API and will showcase some great new games (particularly in the second half of this year.  Also, when you bear in mind the fact that Vista is the only OS to support DX10 you can see that developers and publishers would like to allow a little time for the new OS to “bed in” before they ship games which demand it.  That way there will be a reasonable installed base to appeal to, and for a DX10 game to make commercial sense that’s clearly a requirement.