The high-performance computing (HPC) market is rapidly expanding, and Altera® FPGA coprocessing capabilities are enabling the growth. HPC markets such as financial, medical imaging, bioscience, military, and many others are benefiting from the logic and memory resources in Altera's FPGAs to develop application-specific coprocessors. All HPC markets require coprocessors to provide a performance, productivity, and power advantage. In some cases, Altera FPGA coprocessors accelerate algorithms in excess of 100X. Solutions are now shipping in the following categories:
- Appliances: database acceleration, financial market data acceleration
- Functions: random number generators for financial, 1M-point fast Fourier transform (FFT) for military, test, etc.
- Algorithms: SRCs CARTE, Impulse, AutoESL
For high-bandwidth and low-latency CPU interfaces, partners have Stratix® FPGA-based Opteron HT3.0 solutions, Intel front side bus (FSB) and quick path interconnect (QPI) solutions, and memory DIMM interface solutions. Stratix series FPGAs provide devices with up to 1,087K equivalent logic elements (LEs), 50 Mb of embedded memory, 66 12.5-Gbps backplane-compatible transceivers, 4 28-Gbps high-speed transceivers, and 3,510 18x18 digital signal processing (DSP) blocks.
The newest Stratix series FPGAs, Stratix V FPGAs, provide AMD HT3.0 and Intel QPI 1.1 support (available in Stratix V GX FPGAs). In addition, Stratix V FPGAs provide application-targeted devices for protocols including PCI Express® Gen3, Gen2, and Gen1.
Floating Point
In HPC, floating point is a numerical representation system where a string of digits or bits represents a real number. Applications are increasingly using floating-point data types for more accurate results. The commonly used floating-point applications include:
- Fast Fourier transform (FFT)
- Radar
- Bioscience
- Finite impulse response (FIR)
- Financial options trading
- Matrix math
- Molecular dynamics
- Seismic imaging
- Medical imaging
