Advanced Micro Devices plans to release its new FireStream compute accelerator cards based on the Evergreen graphics processing units (GPUs) shortly. The new devices promise to be something more than just data crunchers.
Although ATI, graphics business unit of Advanced Micro Devices, released its latest family of DirectX 11-compliant Radeon HD 5000 graphics boards well ahead of its arch-rival Nvidia Corp., the latter was quicker to roll out its Tesla 20-series stream computing accelerator cards powered by the latest Fermi-architecture processors, which are now available in servers from Dell and IBM. However, AMD has not been wasting time and plans to introduce the new generation FireStream cards later this month.
Sources close to AMD said that AMD intends to launch “a new AMD FireStream compute accelerator specifically designed for HPC, cloud gaming and server-side rendering, all of which require maximum compute density” by the end of June, 2010, which is nine month after the latest DirectX 11-capable graphics boards for consumers emerged.
Initially AMD FireStream solutions were only designed for a variety of high-performance computing (HPC) applications and the vast majority of such solutions were highly proprietary. However, it is rather noteworthy that the new-generation of AMD FireStream compute processors are also designed for “server-side rendering”, which is something that previous-generation solutions were not intended for.
Actual details about specifications of the new AMD FireStream solutions are not known at the moment, but it is highly likely that they will carry code-named Cypress graphics chips with up to 1600 stream processors.
AMD reportedly plans to promote the new FireStream solutions along with its Maranello (2P, 4P) and San Marino (1P, 2P) server platforms which feature AMD Opteron 6000-series processors with up to twelve cores and AMD Opteron 4000-series chips with up to eight cores, respectively.
AMD did not comment on the news-story.