I'm not sure where to put this, so please mod triage away.
My abridged version:
Modeling heat flow in computer chassis is hard because of the complexity of the flow and the low density of sampling for parameters like temperature. Recent surges in consumer level computational power and the availability of Computational Fluid Dynamic modeling tools such FLUENT make it possible to set up and test the relative performance of such systems with only a few unmeshed heat sources, limited dissipation avenues, and turbulence inducing features such as fan grills.
The paper tests a single case configuration vs. three commercial heat sink designs via modeling- with conclusions that discuss modeling techniques, not case design. However, if you look into the results section of the paper, they do show that one of the three commercial designs is clearly better than the others in modeling (Alpha > CoolerMaster | EverCool).
Hopefully, the more empirically minded folks on spcr might get some ideas from the tools described in the paper and make some really neat toys for us lurkers to gawk at in the future. The link to the original article also mentions a $US 5,000 design contest sponsored by ZeroTherm, if anyone is interested.
Original Article:
http://www.frostytech.com/permalink.cfm?NewsID=68595
Original Paper:
http://www.cse.polyu.edu.hk/publication ... _TariI.pdf
CFD Modeling of Forced Cooling of Computer Chassis
Moderators: NeilBlanchard, Ralf Hutter, sthayashi, Devonavar
That makes intuitive sense, but intuition doesn't always work out in practice.xev wrote:wow that is quite a study. what i understood with a quick glance is that the more surface area the better, and copper cores really help.
I'm not sure that the paper supports any specific conclusions about a particular design- the authors were more concerned with getting the modeling right rather than testing specific designs in a statistically sound manner.