Conjugate heat transfer simulation of cooling of a graphic card
In this project the cooling of a graphics card is recreated using Conjugate heat transfer simulation. Here, the energy equations are turned on and k-epsilon flow model is considered. For the materials of the graphic card copper and aluminium are taken while the cooling medium i.e., fluid domain is taken as air. Copper is designated as the material for the processor(GPU) and the fins, aluminium is used for the remaining parts. A velocity inflow of 10`m/s`is taken with a inlet temperature of 300k. The thermal design power (TDP) which is the amount of thermal power produced by the GPU is taken as 50W to simulate a nVidia Geforce 9400GT graphics card but the model for the graphic card considered here is a basic one. The heat generated is caluated dividing the TDP by the volume of the processor. The volume of the processor is 64`mm^3`or 6.4e-8`m^3` and so the heat generated is found to be 781250000`W/m^3`. First the geometry is changed to a shared topology to get a conformal mesh, the a baseline mesh is created and simulated and after that the mesh is refined to improve the result accuracy. The simulation is carried out until the residuals attained constancy.
The geometry with enclosure:
The geometry without enclosure:
Baseline mesh:
The baseline mesh with enclosure contains 26815 nodes and 147792 elements.
Baseline mesh enclosure:
Baseline mesh metrics:
Convergence plot obtained using baseline mesh:
Baseline mesh temperature plot using mid plane:
Baseline mesh velocity plot using mid plane:
Base mesh wall heat transfer coefficient on the graphic card base:
Base mesh wall heat transfer coefficient on the graphic card fin:
Base mesh wall heat transfer coefficient on the exposed part of the processor view 1:
Base mesh wall heat transfer coefficient on the exposed part of the processor view 2:
Base mesh temperature on base and processor:
Base mesh temperature on base backside:
Base mesh temperature on base and processor showing fin:
Refined mesh:
The baseline mesh is taken and refined. Body sizing is used to refine the mesh created for the 4 cuboidal blocks on the graphics card and edge sizing using number of division method is used to refine the mesh of cylindrical components of the graphics card. The fin is refined seperately by measuring the short edge length using the measuring scale in the display and accordingly the element size is reduced to increase mesh quality. The enclosure is also refined further from the baseline mesh. An inflation layer is added using program controlled inflation which will be helpful in capturing wall heat transfer coefficient more accurately. The refined mesh with enclosure contains 93487 nodes and 504365 elements.
Refined mesh enclosure:
Refined mesh metrics:
Convergence plot obtained using refined mesh:
Refined mesh temperature plot using mid plane:
Refined mesh velocity plot using mid plane:
Refined mesh wall heat transfer coefficient on the graphic card base:
Refined mesh wall heat transfer coefficient on the graphic card fin:
Refined mesh wall heat transfer coefficient on the exposed part of the processor view 1:
Refined mesh wall heat transfer coefficient on the exposed part of the processor view 2:
Refined mesh temperature on base and processor:
Refined mesh temperature on base backside:
Refined mesh temperature on base and processor showing fin:
The maximum temperature attained by the processor is found to be 457.9K. There is only one hotspot in the graphics card, which is the place where the processor is as shown by the above images since processor is the source of the heat generation in the graphics card. Therefore, cooling systems should focus more on cooling the processor part which can improve performance of the graphic card drastically and if possible allow for overclocking to extract more performance.