Conjugate Heat Transfer Analysis on a graphics card.

Aim :

To perform a steady state conjugate heat transfer analysis on a model of Graphics card and to get results based on the behaviour of the heat transfer taking place between the Graphics card and the surrounding air.

Graphic Card Model :

• To simplify the process, let us consider a model with only essential components without any complication.
• The most essential part for heat transfer in the graphics card is the fin. The fins are used to increase the surface area of heat conduction for an effective heat transfer.
• Then we have the source of the heat generated which is the processor of the Graphics card. It is placed just below the fins so that the heat conduction by the fins can be effective.
• Then finally we hav a base which holds upon to the processor & fins. It also contains other small components like transistors, capacitors, Resistors etc...
• By these components our solid phase will be completed.
• To make room for the air, we can generate an enclosure around the solid component.

Fins : Processor :

Base :

Enclosure :

Computational Meshing :

• In this analysis first let us observe the results from the base line mesh and from the results we can Refine the mesh sizing so that we can get more accurate results.
• Thus we will be analysing two different kinds of meshing.

Case 1 :

Element Size :

Mesh :

Case 2 :

Element Size :

Mesh :

Setup & Simulation :

• We will be simulating both the cases with same setup & boundary conditions.
• As we are taking solid body with different components, we will be applying share topology in the spaceclaim wherever needed before meshing.
• In the setup we will bw using pressure based solver & steady state flow.
• Before initializing the boundary conditions, We have to define the materials.
• As the graphics card is made up of components with different materials, here we will be using three different solid materials for the simulation.
• We are considering aluminium as the base material as it is commonly used in computer hardwares. The material properties are taken from Fluent database.
• For the fins we will be using Aluminium alloy A204 which will be suitable for the efficient heat transfer with the air. The material properties are user defined as the alloy is not available in data base.
• Most of the processors are made with silicon because of its properties to withstand the temperature produced. The material properties are user defined.
• After including the materials in the list of solids, we can assign the appropriate materials to the respective components through cell zones.
• After assigning the materials, The source term is specified to the processor which defines the power and heat generated by the processor.
• By considering that the processor uses 10 W of power, we can calculate the source of heat.
• Then we can specify the boundary conditions. here the inlet air is considered to have a velocity of 5m/s.
• As the air undergoes convection with the solid body, we can specify the Heat Transfer Coefficient for the solid body in the Boundary conditions.
• After specifying the values we can start the simulation to find out the heat transfer & Heat distribution.

Contour plot for Case 1:

Refining the mesh :

• From this plot we can get the basic characteristic heat transfer in the sollid phase.
• Using this we can find the critical points where furher accuracy will be required, like the processor need much more finner mesh for the heat generation to be more accurate.
• Considering these issues we can refine the mesh like we did in Case 2.
• We used body sizing for all the components seperatly, so that we can make sure that the total number of elements is under control.
• For critical points and corners, we use smaller element size to increase the calculating accuracy.
• We also refined the enclosure & the proximity setting is enabled, so that we can get finner mesh around the solid bodies. This will give more accurate results around the outer body where the air removes heat.
• Now to compare the variation between the two different meshes, we can check the average temperature of the solid body for both the results.

Case 1 Plot :

Case 2 plot :

Conclusion :

• We can see that the convergence is attained much faster after refining the mesh.
• The average temperature after convergence is also varying from both the cases.
• This means that the base line mesh is not as much accurate than the refined one.
• By using the refined mesh results our predictions can be accurate.

Effect of Varying Velocity :

• Now to see the effect of velocity of the inlet air, we can check the results for different velocities.
• All the setup & Boundary conditions are taken as mentioned before.
• We only change the velocity of the inlet air to see the effects in the result.
• Here we are going to compare three different veloities 1m/s , 2m/s & 5m/s.
• These cases are simulated with the refined mesh.
• We can compare the results of the heat transfer & distribution that occurs in each case.

Velocity 1m/s :

Temperature Contour :

Average Temperature :

Temperature Over Plane :

Velocity 2m/s :

Temperature Contour :

Average Temperature :

Temperature Over Plane :

Velocity 5m/s :

Temperature Contour :

Average Temperature :

Temperature Over Plane :

Conclusion on Varying velocities :

• From the plots it is clear that the temperature decreases as the velocity is increasing.
• Higher velocity air is able to remove heat energy at a faster rate.
• The cooling of the GPU is more efficient when the velocity is high.
• In lower velocity case the temperature over plane plot shows that the air carries the heat for a long time. Where as in case of high velocity the heat drops quickly. This means that the heat convection is effective in high velocity case.
• To remove heat effectively we can use the maximum amount of velocity allowed to the inlet air.
• Thus for further conclusions consider the results of velocity 5m/s.

Maximum Temperature & Heat Transfer Coefficient :

Temperature contour:

Maximum Temperature : 352.1 K

HTC contour:

Maximum Heat Transfer Coefficient : 504.1 W/m^2K.

Convergance :

• The objective is to find out the Heat transfer between the solid & fluid and also the Heat distribution in the solid surface.
• Thus by taking the average temperature on the body we can prove convergence.

• From the plot the temperature attains a steady state after 50 Iterations.
• We can also specify the minimum convergence value for more accurate results.
• As there is no much divergence in the result we can say that the simulation is converged in the solid phase.
• We can also prove the convergemce for fluid phase.
• Consider the temperature of the air that leaves the GPU.
• Create a plane just after the end of the GPU and moniter the average temperature at the plane.

• The fluid temperature also gets to a steady average value at the plane.
• Thus convergence is also attained in the fluid region.

Hotspots :

• With the temperature & HTC contours we can find out the critical points where the possibility of damage is high.
• The HTC value behind the fins is low because of air recirculation.
• The base area beneath the fins & processor got high temperature.
• The fins got a high temperature at the center.