Celoxica (LSE:CXA) presented results of an oil-exploration algorithm customer benchmark at the AMD Torrenza Initiative seminar. The algorithm achieved a 28x performance improvement on a forward wave migration algorithm. At the semimar, Celoxica also presented a financial analysis that showed a consistent 10x return on capital invested on FPGA-enabled servers, and a 50% increase in return on investment by adding one FPGA acceleration card to each server.
Celoxica’s accelerated computing solution consists of an FPGA acceleration board, algorithm acceleration libraries, and a communication interconnect kernel based on the high bandwidth, low latency HyperTransport(TM) technology standard. Celoxica enables FPGA co-processing with their proven programming compiler which allows applications to be accelerated by software developers using a familiar “C” language design flow.
Celoxica delivers an integrated accelerated computing solution, including programming support to enable customers to use a software flow to accelerate their applications. Typical benchmark results for FPGA co-processing have shown from a 12x to 50x performance improvement, depending on the amount of parallelism available in the customer algorithm. To augment these results, Celoxica has developed a calculator that figures accelerated computing return on investment based on hardware purchase and support costs, power savings, reduced cooling costs, reduced space requirements, and improved revenue opportunity from increased computing performance. Customers typically realize 50% higher returns by adding FPGA co-processors to their standard server configurations.