Will ATI Play Nice?

Justin Kerr

Recently in both the print and online versions of Maximum PC we looked at Nvidia’s CUDA API and what a GP-GPU future might look like. The one wild card in this equitation is the other big player in the graphics card market, ATI. Will ATI play nice by supporting CUDA and licensing PhysX? Or will it go its own way, a result which may end up killing both companies initiatives.


Nvidia’s CUDA may not be open source, but it is free. ATI or other companies can easily write apps for the API or implement the functionality into its drivers, but will they? Of course not! The dollar cost may be free, but adopting the format means ATI loses control over the future of GP-GPU computing. ATI openly admits it is currently hard at work with its StreamSDK and Brook+ implementations and its future only takes it further from CUDA. I could write an entire article on the differences between the different approaches but suffice it to say, this is one area where consumers would benefit from a common platform. To prove my point we need only look at the PC gaming market in general. Does anyone actually think we would have been better off with 30 different non Microsoft operating systems and no DirectX? Developers need standards and it is up to hardware manufacturers to make this happen.


Inside sources at Nvidia have also confirmed that ATI can license its PhysX technology for pennies per GPU. ATI on the other hand, has shown no indication that it will do so and ATI has warmed up to the competing Havok solution owned by Intel. As a result the physics scene chaotic as it is will be slow to mature. Choosing between competitors for a physics solution must be no easy task. By siding with Intel, ATI is making a strategic decision that leaves the market divided. In fact, I would speculate that Nvidia’s low licensing fee is an attempt to seed the market with more PhysX compatible hardware to earn a dominate market share.


GP-GPU computing is still a long way out for mainstream users, but it shows incredible promise for many tasks which currently fall under the arduous category such as video transcoding. Additionally, physics acceleration could revolutionize the gaming experience. With these features combined, the graphics card of the future may be far more valuable than it is today. Unfortunately, the revolution may be held up in competing technologies that end up hurting the cause they are fighting for.

Around the web