LATEST INTELLIGENCE
PRESENTED BY
PROTECTING HIGH-FREQUENCY TRADING PERFORMANCE WITH WATERLESS DIRECT-TO-CHIP COOLING
The challenge: cooling risk is business risk
For a high-frequency trading( HFT) firm, compute performance, latency consistency, and uptime all depend on cooling performance. Trading platforms and the modeling workloads behind them depend on processors operating within tight thermal limits, especially when market activity and compute demand spike. If temperatures rise, processors can throttle, and performance can become less predictable. If a conventional water-based cooling system leaks, the result may be damaged equipment, emergency remediation, and unplanned downtime.
HFT organizations already understand what every minute of availability is worth to their business. The question is how to increase cooling capacity without introducing additional risk of failure or adding maintenance burdens at sites that may not have dedicated liquidcooling personnel.
The thermal demands of HFT infrastructure are also changing. Trade execution may remain concentrated on latency- optimized CPUs and FPGAs, but the research, simulation, and predictive-modeling environments supporting moder strategies increasingly incorporate power-intensive CPUs and GPUs. These workloads do not make every trading facility an ultra-high-density environment. They do, however, create concentrated thermal loads that can strain air cooling and reduce the thermal headroom available during periods of peak activity.
Solution at a glance: waterless cooling without added complexity
Two-phase direct-to-chip cooling provides a different approach. Using a dielectric refrigerant instead of water enables trading firms to:
1. Increase cooling capacity 2. Eliminate the risk of water exposure inside the IT environment 3. Support next-generation CPUs and GPUs 4. Simplify operations and reduce maintenance 5. Deploy incrementally within existing facilities •
20
INTELLIGENT CIO NORTH AMERICA www. intelligentcio. com