ATI to take GPUs into the stream age
Enterprise apps to run on games hardware, company hopes
ATI - effectively AMD - has begun pitching its Radeon X1900 graphics chip at corporates, not for running graphics on workstations but for processing large data sets much the way Intel's Streaming SIMD Extensions (SSE) are meant to.
Dubbed 'Stream Computing', ATI's initiative pitches the technology at scientific and engineering applications such as oil-field analysis and chemical experimentation, though ATI was quick to admit the system may also find a home in more consumer-centric applications too, and in mainstream business for database searching and financial modelling.
While not explicitly associated with games physics processing, Stream Computing works much the same way: specially designed coding tools allow the GPU's many pixel shader processors to crunch through non-graphics applications. It's all bits and bytes after all, and it's up to the host application to interpret the results the GPU churns out as hydrocarbon pressure densities rather than exploding Strogg.
The key is that such applications involve large blocks of data, all of which require processing in the same way. So-called vector information has been targeted with dedicated processing tools in mainstream CPUs, with systems like Intel's SSE, AMD's SSE-aping 3DNow! and Motorola/Freescale's AltiVec.
There's a way to go, and for now ATI is simply punting the concept and encouraging corporates to explore ways in which its GPUs could accelerate their calculations. Enterprisises need to tie their existing and future apps into ATI's GPUs, and that means working either with ATI or with third-party API designers who'll provide the middleware to bridge enterprise application and GPU. Together, they'll also need to develop code to allow racks of GPUs to co-operate on single applications. ®