Modeled the machine in AutoCAD
Posted by John-Erik | Filed under Uncategorized
I modeled the machine in AutoCAD completely and then had several different cable and fan arrangements. I also had the nominal heat outputs programed in, as well as minimal and maximal. I then uploaded it to one of the Suns with a customized fluid flow package and ran it. It took about two to three hours, but I got an output of both the best cable placement. and best fan placement and minimal, maximal and normal temperatures I could expect. Although this was for my old CPUs, my heat output hasn’t increased dramatically enough to recalibrate it. The one I picked had a lower maximum heat than any other design’s minimum heat, so it was a no brainer (and this was the one I suspected would be best as well, but I wanted to make sure and I’ve been given way too much HPC time in the past than is good for me).
The package front ends are custom written but use one of the two or three standard fluid flow modeling cores (remember, there really aren’t that many out there). It does not interface directly to AutoCAD but can take an AutoCAD model as input, and since most of the items were simple shapes with protrusions or cutouts it was fairly simple to create multiple models, and since I also have many repeating shapes it was again trivial. It took me about three days to model the system and then I uploaded it.
The reason it took so little time is I was uploading it to a Fire V880 server with both the RAM and CPU time to complete it quickly. It has roughly four times the power of the Crays that were used to model entire submarines in the late 80’s and early 90’s, and those jobs took only a few minuites.
I could determine it by emperical measure, but I didn’t have extra side panels to try it out on. However, given how low my temps are I doubt that I chose the wrong arrangement either.
Similar Posts
- How to tell poor thermal transfer
- 80mm fan in the top
- Conductivity and Corrosion
- Power Supplies Compared
- New ‘Old’ Videos Posted
