Markets
ClearSpeed Technology's Advance accelerator product family has been specifically designed for the unique challenges faced by the most demanding users of numerically intensive computing. In markets such as the financial services industry, top universities and national laboratories the relentless demand for compute cycles is in direct contention with the constraints of facilities space and the limits of power and cooling.
Solutions for the financial services industry
Getting to market early provides a competitive edge and wins business. For quants and traders more accurate valuations and quicker risking informs trading decisions. A lack of available compute capability and lengthy development times can be a significant obstacle to achieving an advantageous time to market. Some instruments are just too costly or time consuming to price and risk using existing technology. The availability of high performance, low power, scalable acceleration technology now provides a solution without requiring datacenter upgrades.
Performance comparisons based on benchmark code for European Option pricing provided by a major international bank showed up to 20x performance speedup using a ClearSpeed Advance accelerator compared with a typical industry server. The use of multiple Advance accelerators in the same system delivered up to 100x performance speedup.
Solutions for Universities and National Laboratories in science and research
The vast majority of applications in science and research depend upon the solution of a dense system of linear equations which in turn depend upon a system's performance and accuracy with double precision floating point calculations. ClearSpeed's massively parallel CSX600 architecture has been specifically designed to excel in the most numerically intense applications whose requirements are frequently based upon linear algebra performance characterized by the LINPACK benchmark.
High Performance Computing (HPC) Markets
Despite its inherent diversity it is possible characterize much of the HPC market into a manageable number of segments. The International Data Corporation (IDC), which tracks the HPC market, defines seventeen separate significant application categories. It is possible to simply this further and to map them to the following diagram:
While some HPC applications are integer oriented, some floating-point dependent and some a combination, most exhibit significant levels of data parallelism making them suitable candidates for acceleration.