Wall Street's thirst for faster processing speeds and low-latency analytic computations is turning into a game, of sorts. Graphics cards that are used to power game consoles such as PlayStation and Xbox are turning up on quantitative trading desks, where they're being used to calculate options analytics.
Nvidia, the Santa Clara, Calif.-based graphics chip maker that rode the video game boom and the demand for 3-D graphics rendering, is producing high-performance computing appliances based on graphics processing units, or GPUs, that are much more powerful than conventional hardware. Nvidia has sold the GPU processors to the oil and gas industries, and universities doing research on cancer or molecular dynamics, according to the company. With compute-intensive applications such as real-time options valuation, Wall Street now is eyeing the technology as well.
"With the advancement of computer gaming and graphics technology like PlayStation and Xbox, these graphics cards are a lot more capable [than ever]," says Frank Tan, an options trader at Deutsche Bank. "The idea is to use the GPUs for financial calculations rather than playing games and generating graphics."
According to Tan, investment banks face two main processing challenges in the options market. "First, there is the sheer number of messages and data coming in," he says, noting that most banks' networks can handle the message volume. But options trading desks are tackling a second problem, Tan adds.
Every time an equity or option ticks on an exchange, traders need to recalculate all the options contracts associated with that underlying data, Tan explains. Microsoft alone, he points out, has more than 150 options contracts, and there are approximately 300,000 (and growing) exchange-traded options contracts.
"Every time the underlying price changes, or if there is a shift in the options quotes, the options prices and the Greeks have to be recomputed," Tan says. "The GPU has the capability to solve the second computational issue. Before GPU technology came out, in order to solve this problem in real time, you would have to buy supercomputers."
According to Jerry Hanweck, founder of Hanweck Associates, a quantitative financial consulting firm in lower Manhattan, GPUs offer more power at lower costs than traditional CPUs. "The reason this becomes interesting for options is that the message rates are increasing, which means that traders who want to keep up with that message traffic and do these kinds of analytical calculations need to do this faster," he says, cautioning that firms that don't jump on the GPU bandwagon will need to buy a huge number of computers.
"They've got to power them and cool them and find space to put them in," Hanweck notes, adding that rack space on Wall Street is costly. "It gets expensive."