Automated Gimbaling of Conical Nozzles for Jet Vectoring Impingement Cooling to Mitigate Workload Dependent Hotspots in High Power Density Silicon

Direct jet impingement on the surface of the bare die silicon delivers lot more ability to remove heat instantly because of the higher heat transfer coefficients that impinging jets can deliver. High power density silicon dissipates higher heat loads especially in non-uniform power distribution on the silicon in different regions within the die. On a non-uniformly powered die, power tends to be distributed and concentrated on certain areas only thereby driving higher junction temperature (Tj) rise resulting in localized hotspots on the die. Differing workloads produce different distinct hotspots in different regions of the silicon. We introduce a novel approach for improving heat dissipation on high-power density silicon dies even with non-uniform power distribution. The proposed technique involves automatic gimballing of conical nozzles to "jet-vector impinge" on hotspots targeting localized areas of the die experiencing higher power dissipation. Controlled by a micro-controller and a closed-loop algorithm using dynamic temperature feedback, this method continuously mitigates hotspots on the silicon, reducing overall junction tempera ture. The gimballing or jet-vectoring is done based on sensor feedback, ensuring multiple nozzles target different hot zones within the die and is done dynamically. For client silicon, a total of 30 distinct CFD RANS simulations of turbulently impinging jets issued from a single circular nozzle and accelerated under an orifice plate comprising sixteen tapered micro-scaled nozzles in confined channels at different nozzle to silicon wall separation (8 ≤ z/d ≤ 24) distances at angles of impingement, ∅, from 30° (oblique) to 90° (orthogonal), are examined using Ansys Icepak. 15 cases of orthogonal impingement and 15 cases of jet vector impingement on hotspots were simulated. For multi-chip modules, a total of nine RANS cases were investigated, three CFD cases before jet vectoring and six cases after jet vectoring were investigated at angles of impingement, ∅, from 55° to 90°. Application of jet vectoring technique demonstrates a notable reduction of junction temperature by 10℃ for a monolith client silicon and 20℃ for multi-chip modules with higher power density, at identical boundary conditions before jet vectoring, presenting promising results for enhanced heat transfer and more uniform temperature gradien ts across the silicon surface. Results from Reynolds-averaged Navier-Stokes (RANS) computational fluid dynamics simulations are provided and discussed for both pre and post jet-vectoring impingement scenarios.