With the semiconductor process technology promptly following Moore’s law and with no great changes to the fundamentals of processor architecture, CPU designers have resorted to multi-core and many-core architectures to make full use of available semiconductor real-estate. The advent of multi-core and many-core processors is enabling the scientific community with tremendous processing power at affordable price points. Though hardware has advanced rapidly, software is lagging quite a bit in exploiting the capabilities of modern day CPUs.
Since many applications are serial in nature, there is tremendous scope for parallelization and optimization of HPC applications (aka Code Modernization) to bring in several orders of performance improvement. Once modernized, benefits of code modernization will carry forward to the next generation CPUs seamlessly. The biggest advantage of Code Modernization is fast turn-around-time of results. This means more runs, better results, high quality research. And the best return on investment!
Recently, FPGA-based acceleration using add-In-cards is gaining lot of momentum in increasing performance of applications running on Servers / HPC clusters. Instead of pure programming, lot of performance is being achieved by executing part of software on acceleration hardware. Reprogrammable feature of FPGAs is making it an ideal choice for performance enhancement of applications running on HPC cloud.
Some key services offered by Cabot Partners in HPC/HPDS/HPDA are as shown below: