The Gateway to Algorithmic and Automated Trading

BittWare announces TeraBox multi-FPGA system

First Published 10th September 2013

BittWare's high-performance FPGA platform provides high density network processing and high performance computing applications.

Waltham, Ma - BittWare, the provider of Altera-based COTS FPGA boards, has announced the availability of the TeraBox, a multi-FPGA system for high performance computing and network processing applications.

The TeraBox provides up to sixteen Altera Stratix V FPGAs in a single system up to sixteen, for 20 TeraFLOPS of total processing power, along with 6.5 Terabits/sec of memory bandwidth and 1.28 Terabits/sec of I/O - all in a turnkey rackmount solution.

The system features up to eight full-size dual Altera Stratix V GX/GS PCIe boards - BittWare's S5-PCIe-DS (S5PE-DS). Along with two Altera Stratix V FPGAs per board, the S5PE-DS also provides an extremely flexibly memory configuration with 8 SODIMM sites each supporting DDR3 SDRAM, RLDRAM3, or QDRII+. The TeraBox supports up to a system total of 512 GBytes spread over 64 banks for 6.5 Terabits/sec of memory bandwidth. The system arrives fully tested and configured, and includes complete development software support with BittWare's BittWorks II Toolkit, allowing users to immediately focus on their specific application development.

"Our S5PE-DS card is the most capable FPGA processing board available to date on a PCIe format. Putting up to eight of these boards into one rackmount chassis, we have created an extremely powerful, highly flexible system," stated Ron Huizen, Vice President of Systems & Solutions at BittWare.

"The off-board I/O options of the S5PE-DS enable low-latency interfacing and integration to a wide range of systems and networks which will have seamless access to sixteen Stratix V FPGAs and a large amount of on-board memory. The system is ideal for high performance computing and high density network processing applications that require an obscene amount of FPGA processing and comes instantly ready for high performance computing application development."