
Conventional air cooling moves heat by pushing more air over more fin area. JouleForce changes the interaction itself, using airflow through patented microchannel arrays to extract heat inside the channel, at the wall. JouleForce is unimpaired by boundary layer formation that is the limiting factor of conventional air cooling.
Shown above: the microchannel array at macro scale, the surface where the physics happens.
"This effectively extends the practical limits and lifecycle of air-cooling technologies."
Dr. Ortega reproduced Forced Physics' thermal-resistance and pressure-drop results within 3-5%. Dr. Ortega's evaluation also found 50% lower thermal resistance and one-third the airflow vs. the Dell reference heat sink.
Villanova reproduction of Forced Physics' SP5 results.
At one-third the airflow of the Dell reference heat sink.
A conventional heat sink relies on boundary-layer-laden bulk airflow sweeping across fin surfaces. As chip power climbs, the only levers are more fin area, more air, and more noise. JouleForce works differently.
Heat transfers only where moving air contacts the fin. A boundary layer of slow, near-wall air limits how much heat actually leaves the surface.
Heat transfers along the full length of each channel wall. The geometry diminishes the boundary layer, permitting more air to come into direct contact with the heated channel walls, so more heat leaves with less airflow.
A fan or blower draws air through the microchannel array by maintaining a pressure differential through the device. Airflow is required, but at a fraction of the volume conventional cooling demands.
The microchannel array is mounted to a thermal interface, such as a vapor chamber, that transfers heat from the device being cooled directly into the microchannel surfaces.
Geometry-driven molecular collisions extract heat from the microchannel walls more effectively than classical fluid dynamics predicts. The boundary layer that limits conventional air cooling never fully forms, keeping the heat exchange mechanism active along the entire microchannel length, and allowing the air to absorb more heat.
The warmed air exits through a guided exhaust path and the cycle repeats, sustaining low chip temperatures under constant load with less airflow than conventional cooling requires.
JFA · microchannel detail
The performance comes from the array geometry itself: the channel dimensions, spacing and flow path. This is what allows JouleForce to increase heat removal without relying primarily on more airflow or larger conventional fins.
Internal testing on the NVIDIA RTX 6000 showed substantially lower GPU junction temperature and improved sustained token throughput compared to the stock heat sink under the same workload.
Higher token throughput under sustained AI inference load, NVIDIA RTX 6000.
24hrs, 7 days a week, everyday of the year. A JouleForce-cooled edge system operates continuously in Phoenix, Arizona, using only outside air to cool the electronics. CPU temperatures remain within a tight band as inlet conditions change, with no recorded throttling events during the representative period shown.
Representative field telemetry from a JouleForce edge deployment. Full test conditions and logs available under NDA.
The core array geometry and its implementation are protected by an issued patent portfolio. That matters because the advantage is not a manufacturing trick that can be copied, it is the geometry itself, and the geometry is owned.
U.S. Patent 10,379,582 B2 · “Assembly and Method for Cooling” · 14 claims, 32 drawing sheets · Forced Physics LLC.
Independent reproduction of Forced Physics' SP5 thermal-resistance and pressure-drop measurements, with comparison to conventional air cooling.
Download the independent evaluation →A deeper explanation of the microchannel architecture, heat-transfer mechanism, test methodology, and design implications.
Download the engineering whitepaper →
Tell us the target hardware, power envelope, ambient condition, and deployment environment. We will scope an evaluation around measurable thermal and performance targets, and share the full methodology under NDA.