C.H.T Analysis on a Graphic card using ANSYS FLUENT

I. Introduction :

In this project, A steady state conjugate heat transfer analysis on a Graphic card model is done. Graphic card has become an everyday used object and a very importat part of any computer system, laptops etc. This product is mass produced daily in millions and has made computers exceptionally efficient. As they consume electricity to operate and running a heavy software or graphically challenging task on computer takes a toll on its graphics card, because of which it generates a large amount of heat. There is a limit to which any system can entertain heat to keep its parts under safe operating condition. We cannot go past these limits as it will damage the system, Cause failure of the product & system etc. But what we can do is find measures to cool this system, cool the graphic card so the system keeps running at optimum speed & condition. 

There are various methods of cooling electrical components, through fan, thermal pastes / gels, double fanning, using materials that help dissipating the heat in the fans direction, arrangements of components etc. We have a graphic card model here which we will simulate in ansys fluent by applying heat on the processor/gpu of the graphic card while its contained in an enclosure. The air is passed at a certain velocity through the enclosure and we observe the heat dissipation through the fins by conduction from the processor, then observe how the flowing air cools the fin & observe the heat transfer taking place through it.

II. Parts & Materials of Graphic card :

There are many parts in a graphic card namely, Gpu/processor, fins, heat sinks, pcbs, connecting copper coiled wires, memory modules,pci connection port, connector pins etc.

The model used here is simplified to 3-4 important parts and isnt heavily detailed to save computational time & prove the point through analysis.  

The parts of simplified model are - Processor, Fin & base. small memory modules are attached in as base. 

Materials Considered for the analysis . 

For processor - Silicon is choosen for this. Silicon is not the only material of the processor, its a very complex part made of mixtures of materials, for convenience, knowing silicon forms one of the major part of it its choosen. 

Fin - Aluminium . 

Base - PCB 

Fluid - Air. 

Material Properties defined :                                                                                                     

III. Objectives : 

1. To conclude how the simulation has reached convergence.

2. Maximum temperature attained by the processor. 

3. Htc (heat transfer coefficient) at appropriate areas of the model.

4. Potential Hotspots on the model. 

IV. Project Approach : 

Case 1A - This case is solved with a baseline mesh with a minimal velocity inlet of air .  

Case 2A - With exactly same mesh & conditions, velocity of air is increased to perform a short case                      study to observe the changes in the max temperature of the processor. 

Case B -  Mesh refinement is done and Setup of Case 2A is followed i.e its velocity. 

V. Geometry Preparation in Ansys spaceclaim : 

Case 1A, 2A & B

An enclosure is created around the Graphics card body.

Share prep is enabled to share the data between the enclosure and the bodies of graphic card.

VI. Meshing : 

Mesh data : Mesh data is a tabulated form of the generated mesh shown below after this on case basis.

Case 1A & 2A:   

This is a baseline mesh without any refinements.

Case B : Body sizings were placed at proper bodies & inflation layers created around it. 

It was important to create an inflation layer around the body to refine the solution in as much detail as possible.  As these outler layers capture the heat being transferred i.e convection phenomenon from solid to fluid.  Body sizings are assigned to all solid bodies. 

VII. Case Setup : 

1. Pressure Based Solver. 

2. Energy eqn on 

3. Viscous model - k eps 2 eqn.

4. Time : Steady

5. Cell zone - Material set.

6. B.C - 

    Case 1A - inlet : 3 m/s 

    Case 2A & B inlet : 8 m/s

    Inlet & outlet temp : 293.15 k

    Pressure outlet 0 pascal. 

7. Source Term in Processor : 154687500 w/m^3. 

To consider for heat generation in the processor many standard values were referred to from the internet & catelogues of Graphic card manufacturing companies like Nvidia. The power consumption lies anywhere between 40 to 120 watts from low end to high. However our gpu model's size is a little smaller than standard ( refer in the reference) . So the power Consumption considered is 9.9 watts. 

The volume of model processor is 64 mm^3. 

Hence calculating the Source Term as such : 

Source Term = `9.9/(64*10^-9) = 154687500 W/m^3` 

VIII. Solution & Post Processing Results : 

Case 1A & 2A

Residual Plot :  

Case ran for 200 iteratios in which all the equations have stabilized at around 140 iterations. 

Maximum Temperature Attained by the Processor:

We can observe that for case 1a with the velocity at inlet being 3 m/s the maximum temperature attained by the processor is 169.8 degree celcius. While the temperature attained during case 2a is 123.392 degree celcius with the inlet velocity of air as 8 m/s . 

This small case study concludes that the heat transfer is occuring at a faster rate where the air velocity is higher given which it cooled the Graphic card more. 

Case B :

Residuals :  

The case was also ran for 200 iterations which converges around near 120 iterations. 

Maximum Temperature Attained by the Processor:

We have much finer mesh in this case about 1.4 million elements.  The case setup is exactly same as Case 2A . while the case 2A's maximum temperature at processor came out to be 123, with a highly refined mesh it gives around 84 degree celcius. 

More over we can observe the The air being stagnant at some parts of the fin and air also gets into a recirculation zone, heat transfer from these areas is lower and thier htc is low as well. 

The velocity plot shows the air velocity at 8 m/s 

The vector lines show the recirculation zones and stagnation zones a bit more easily with arrows.

IX. Heat transfer coefficients :

The most high heat transfer rate in a graphic card is occuring in the fins, its logical. So upon comparing the Cases we observing for low velocity inlet we have less heat transfer coefficient and i.e less heat transfer while increase for increase in inlet velocity. 

We can observe that from the more refined case B, the area where the air is stagnant the HTC is significantly lower than the tip of the fin.  

For case 1a it is 401.7 w/m2k & B it is 2030.6 w/m2k. 

 In the base the heat transfer coefficient increase for case 1A showing 964 w/2k at the memory module region.  Similarly in the Case B too and its around 1015.3 w/m2k . 

 This is the generalized wall htc on the central plane reigion , the red zones irregular patterns show the wall htc of those areas for Case 1A. 

X. Potential Hot spots on the model : 

The pcb is the hot spot as the processor is directly attached over it. This picture shows front and back of it. 

 Other ones are fin and the processor itself. The central region of the fin more than its surroundings as the processor is near its centre.  

Mesh view of the Case 1A . 

   Animation : 

XI. Conclusions : 

1. We conclude the solution convergence by first observing the residuals where all the equations converged . Secondly the desired results where achieved as we expected it. i.e the idea in mind of what the cht simulation results would prove was infact observed in every iteration of the solution which leads us to conclude a fully converged solution. 

2. The max temperature attained by the processor will always vary, like it increased when the air flow was less and decreased when it was more. hence proper cooling of such electronic devices is necessary for efficient functioning.  

3. The conjugate heat transfer study allowed us to peak into electronics heat distribution, working and many conceptual clearance which are before theoretically verified in hand calculations or text book principals. 

4. Heat transfer coefficient is something which changes with materials used, hence engineers are very picky about the materials used in these cards as they are so compact, the need to provide higher heat dissipation and improving performance directly results in earning more from customers. Thus heat sinks like fins, other shorter heat extending devices etc are carefully manufactured for this very purpose. 

5 CHT study helps in identifying hotspots of the model in thermal analysis, these hotspots once discovered, gives the engineer the ability to make changes in the model where necessary and improve its thermal efficiency and performance. 

6. Proper meshing has extremely important advantages as we observed from case 2a and B having significant differences in terms of maximum temperature. 

7. Hence having an expertise in software simulations, meshing in particular and knowledge of models required to perform these simulations can help develop a better product. 

XII. Reference :

1. https://www.techpowerup.com/forums/threads/major-nvidia-die-size-comparison.247247/

2. https://projects.skill-lync.com/projects/CHT-Analysis-on-an-Exhaust-Manifold-52149

3. https://en.wikipedia.org/wiki/Graphics_processing_unit

4. https://www.pcgamer.com/best-gpu-2016/

5. https://www.reddit.com/r/nvidia/comments/9bwihc/major_nvidia_die_size_comparison/

keywords - ANSYS-FLUENT, CFD, GRAPHIC-CARD, CONVERGENCE STUDY, CHT 

   ------------------------------------------- THE END -------------------------------------------------------