Berkeley CIFT 

Market data flows are voluminous, fast, time-critical, and growing rapidly around the world.

There are particular challenges to meet when designing computers for electronic trading, and a particular aspects of the data and analytic task that suggest solutions using current technology.

rs, and now are talking about microseconds. Many installations are careful about cable length. Latency that would be unnoticeable or immaterial in other contexts is fatal in moving market and algorithmic trading data.

Computation of analytics and network overhead are also time consuming, and therefore, for the fast end of the financial market, money-losing concerns.

We have seen this before. Audio computed on the CPU stuttered. Graphics on the CPU was fine for Pong, not so fine for photorealistic games and video.
The solutions to these problems, specialized sound and video hardware, has spawned a significant subindustry. Creative Labs Soundblaster and successors, Nvidia & ATI display cards.

This worked because there was general agreement on what this specialized processors should do, expressed in standards like DirectX that allowed developers to use the hardware. The hardware itself could exploit the inherent parallelism in the processes, particularly in graphics, where, for example, calculating the reflections from a light source on an object in one part of a scene is fully independent of the calculation in another part. t

There is an inherent parallelism there which is also present in many financial trading applications. Analytics computed on the a limit order book are computed on a single stock, and there are roughly 40,000 investment quality stocks trading globally.
8 cores are nice, but this is a problem that could benefit from 800 or 8,000 specialized silicon engines.

At the lower levels all communications protocols are done in hardware, software at the top. Market application protocols, for order and execution management want to live in hardware.

This could be a significant opportunity for chip and hardware firms. Financial markets now rival defense as the leading edge users of advanced computation. To recycle one of my favorite titles, "If you have everything computationally, where would you put it financially?"