Conjugate Heat Transfer Analysis on a graphics card.
Aim To conduct a steady-state Conjugate heat transfer (CHT) analysis of air-cooled Graphics card using CFD simulation in ANSYS FLUENT and observe the following:
Solution convergence
Temperature distribution
Identify hotspots (regions of maximum temperature)
Heat transfer coefficient variation
Effect of air velocity on the cooling effect.
Introduction
Graphics card is computer hardware that produces visual output, dedicated graphics card are used to enhance the visual quality involving very high computing power. These are commonly used for gaming, animation, video editing, or any similar activity involving high graphics. To produce such high graphics involves a lot of processing power which in turn consumes high energy as a consequence of which a large amount of heat is generated which increases temperature of the graphics card. Graphics cards work efficiently and safely at a certain range of temperature so a cooling mechanism is required to keep away the heat produced at the processor and maintain the required temperature range else there can be melting of components or short-circuits leading to a breakdown. To increase cooling of the processor through convection, fins are used which increases the surface area available for convection, and generally, a fan is mounted in the cabinet to maintain a proper airflow circulation and to increase the air velocity (forced convection), therefore, increasing the heat transfer coefficient which will increase the heat transfer from the processor to surrounding and thus produce the cooling effect maintaining the required temperature.
Here we consider a model replicating a standard case of graphics card which contains a base plate on which a processor is mounted over which a fin setup is attached to increase heat dissipation, some other components are also mounted on the base plate. Here heat is generated in the processor and transferred to various components and the air is circulated which takes up the heat. The model is contained in an enclosure which will represent the computational fluid domain over which the airflow is considered velocity inlet condition is used at one face to maintain the air circulation velocity.
Geometry
This is the model used representing a graphics card in a very general form containing a processor, fin, base plate with some other components. This model is enclosed in an enclosure as shown below.
Mesh
Two mesh cases have been considered
(1) Base/default mesh for base-line setup
(2) Refined mesh
Section plane is used to show mesh near the interface between fluid and solid material.
Share topology is used in SpaceClaim for conformal meshing at the solid-fluid interface.
Default mesh
Element size = 2mm
Number of elements =318642
Number of nodes = 60737
The element quality plot shows the majority of the elements in the mesh are of high quality with an average value of 0.83.
Refined mesh
Body sizing is used to refine mesh near fins, processor, and other walls to account for better results at solid-fluid interaction for heat transfer coefficient and max temperature.
Material considered
Processor - Silicon
Fin & Other Components- Aluminum
Base-plate - PCB (FR4)
Fluid - Air
Setup in FLUENT
Steady-state analysis
Pressure based solver
The energy equation is used as we want to solve for temperature
Viscous model - k-epsilon
Boundary conditions
Inlet velocity - 1m/s , 3m/s, 5m/s
Outlet pressure - Gauge pressure=0
Symmetry - Rest 4 faces of the enclosure
All parts of the graphics card are assigned as solid zones
Processor - A heat source
considering 30W power consumption by the Graphics card
the volume of the processor in the model considered = 64mm^3
Heat generation per unit volume = 468750000 W/mm^3
Rest components are considered as wall with the coupled thermal conditions.
Solution
Case-1 Velocity= 1m/s
Residual plot to check for convergence
The plot shows solution has converged after 350 iterations.
Temperature contour represents the hotspot at 1881K on the fin plate above the processor area due to heat transfer from the processor to the fin plate.
Due to coarser mesh in this case & as the heat transfer coefficient is calculated in the first cell near to the wall. A vague plot can be seen.
Refined case
The convergence of solution at 260 iterations
Due to mesh refinement heat diffusion and conduction is properly captured and max temperature is found to be 1070K.
The processor is found to be at the highest temperature as it is the heat source.
Base plate Temperature distribution is correct as the area under the processor shows the max temp region, due to heat transfer from the processor to the base plate.
Wall heat transfer coefficient in this refined mesh is properly plotted as the cells near the wall are refined and max value of approx 800W/m^2/K can be seen at the interface.
Due to mesh refinement, a better plot is seen for the heat transfer coefficient. More mesh refinement can give better result but here we were confined to academic version mesh limit.
As seen Max heat transfer coefficient takes at the processor as it is the heat source also.
Similar plots are shown below for different velocities and concluded at last.
Case- 2 Velocity = 3m/s
Baseline setup vs Refined mesh setup
Max temp = 1311K Max temp = 671.1K
Max heat transfer coefficient = 1195W/m^2/K Max heat transfer coefficient = 800.2W/m^2/K
Case-3 Velocity = 5m/s
Baseline setup vs Refined setup
Max temp 1161K Max temp 563.7K
Max heat transfer coefficient =1195W/m^2/K Max heat transfer coefficient = 800.2W/m^2/K
Conclusion
1) Max temp decreases with increase in velocity
2) Max geat transfer coefficient increases with increase in velocity
3) Mesh refinement gives better results for both Temperature and Heat coefficient plots.
4) Hotspots are regions around processor, area above & below processor of fins and base plate
5) Convergence of the solution is confirmed with repeating pattern seen in residual plots
6) Proper material selection can be done considering max temperature attained for better efficiency