AMD's new Firestream chip tops 1 teraflop
GPGPU smackdown week
The latest round of graphics card dueling between AMD and Nvidia isn't just over high-end gamers. The vendors will also exchange blows for the hearts and wallets of your friendly neighborhood medical imagers, seismic modelers, and computational fluid dynamicists.
AMD is refreshing the FireStream processor line with a new general purpose GPU (GPGPU) that boasts more than one teraflop of processing power.
Instead of handling gaming or graphics operatins, GPGPUs are built to crunch hundreds of parallel calculations per clock cycle. They promise massive speed improvements over a CPU in mathematical workloads of the scientific, educational and high performance computing variety.
AMD claims developers are reporting up to a 55x performance increase on financial analysis codes as compared to using a CPU alone.
AMD's new card succeeds the FireStream 9170 , released last November and capable of 500 gigaflops for single precision performance. The FireStream 9250 is capable of 1 teraflop for single precision calculations, or 1 trillion floating-point operations per second.
Nvidia is similarly rolling out its 240-core Tesla-10 Series  GPGPU chip that's also capable of 1 Teraflop of computational muscle.
The FireStream 9250 comes with 1GB of GDDR3 (Graphics Double Data Rate 3) memory — which is actually half of what was previously offered in the FireStream 9170. The 9250 includes double-precision floating point hardware performing at more than 200 gigaflops.
The 9250 fits in a single PCI slot, making it compatible with most desktops, workstations, and larger servers. It's power consumption is about the same as its predecessor, demanding less than 150 watts and a power efficiency rate of eight gigaflops per watt. That's in line with Nvidia's PCIe slotted device that eats up about 160 watts.
AMD plans to ship the FireStream 9250 and accompanying development kit in Q3 for $999. ®