About a-team Marketing Services
The knowledge platform for the financial technology industry
The knowledge platform for the financial technology industry

A-Team Insight Blogs

BittWare’s TeraBox Bulks Up FPGA Processing for Trading Scale and Analytics

Subscribe to our newsletter

FPGA specialist BittWare has introduced TeraBox, an appliance that delivers up to 16 FPGAs, and targeted at high scale trading and analytics applications.

TeraBox supports up to eight BittWare S5-PCIe-DS cards, each of which hosts two Altera Stratix V FPGAs, 64 gigabytes of RAM and 16 10gE network ports. Thus, each 5U appliance can scale to 16 FPGAs, 512GB of RAM and 128 10gE ports. The appliance can also optionally host a traditional x86 processor, perhaps for monitoring or co-ordination functionality.

According to BittWare’s vice president of systems and solutions Ron Huizen, TeraBox has two likely applications in the financial markets:

* For trading systems where the entire application logic is hosted on the FPGA card, TeraBox offers high scale in one appliance, thus reducing cost of deployment compared with server-hosted approaches. Algorithmic trading and real-time risk control are applications that can likely be deployed more cost effectively with TeraBox.
* For analytics applications, such as algo back testing, pre-trade analytics and risk managemnt, TeraBox’s multiple FPGAs can work together to provide parallelised performance. Connectivity between the FPGAs can be acheived via the chassis PCIe bus, or via the 10gE network.

The latter analytics example would be to some extent breaking new ground for FPGAs, which typically have been dedicated to specific functions, such as data feed processing. For the most part, parallel analytics applications have been targeted at multi-core x86 processors, or at GPUs.

A key to deploying FPGAs for parallel applications will be the introduction of the next version of the OpenCL programming framework. OpenCL 2.0 – the specification for which was released this July – calls for support for dynamic parallelism and shared virtual memory.

Subscribe to our newsletter

Related content

WEBINAR

Upcoming Webinar: Reviewing the Latency Landscape and the Next Generation of Ultra-Low Latency Infrastructure

Date: 17 September 2026 Time: 10:00am ET / 3:00pm London / 4:00pm CET Duration: 50 minutes Ultra-low latency is no longer the preserve of a handful of proprietary trading firms. As new asset classes electronify, data volumes surge, and regulatory expectations around execution quality and resilience tighten, the performance demands on trading infrastructure are broadening...

BLOG

DiffusionData Targets Agentic AI in Finance with New MCP Server

Data technology firm DiffusionData has released an open-source server designed to connect Large Language Models (LLMs) with real-time data streams, aiming to facilitate the development of Agentic AI in financial services. The new Diffusion MCP Server uses the Model Context Protocol (MCP), an open standard for AI models to interact with external tools and data...

EVENT

TradingTech Summit London

Now in its 15th year the TradingTech Summit London brings together the European trading technology capital markets industry and examines the latest changes and innovations in trading technology and explores how technology is being deployed to create an edge in sell side and buy side capital markets financial institutions.

GUIDE

The Reference Data Utility Handbook

The potential of a reference data utility model has been discussed for many years, and while early implementations failed to gain traction, the model has now come of age as financial institutions look for new data management models that can solve the challenges of operational cost reduction, improved data quality and regulatory compliance. The multi-tenanted...