Aw hell, here goes…
Graphics hardware prefers models that can be grouped into ordered strips of triangles. The bigger the strip the better. If the model can be broken down into large strips of triangles the graphics hardware has an easier time manipulating it.
As a very simple example, lets say you have a simple ribbon of ten triangles. If you were to break the ribbon up into discrete triangles (similar to what you’ve done to this fridge), the graphics hardware would have to do 3 transforms per triangle per frame. The points that make up the traingles of the ribbon would look like this to your graphics hardware:
1 2 3
4 5 6
7 8 9
10 11 12
13 14 15
16 17 18
19 20 21
22 23 24
25 26 27
28 29 30
This is what’s known as a triangle list. That’s 30 points that need to be transformed per frame.
Now take the same ribbon as a continuous ten triangle tri-strip. Here’s what the graphics hardware sees.
1 2 3 4 5 6 7 8 9 10
Since the hardware takes this strip in order, it has 3 points to transform for the first polygon and only one point for each successive polygon in the strip, a total of 10 transformations as compared to 30 tranforms for the triangle list above. The objects are the same, but this tristripped version is much less work to draw. Since the hardware sees these triangles as continuous and ordered it only has to do one transform for each successive triangle because the other two points that make up the current triangle were already calculated for the previous triangle. So much less work has to be done by the GPU with the tristripped model that we can double the number of polygons in the tristrp version and, in a perfect world, it will still be transformed by the hardware 33 percent faster than the triangle list version of the ribbon above.
This example is small and doesn’t seem like much, but imagine a game where you have millions of triangles to churn through 60 times a second… Saving dozens or hundreds of calculations per object per frame really adds up.
Hopefully I’ve explained this clearly and you can understand that by keeping the surface continuous, even though the object has a couple dozen more triangles, it may draw faster because larger strips of triangles can be created. Your 78 polygon fridge as a solid object, would be composed of several strips, but could possibly be handled by the hardware as easily or possibly easier than your simplified version of the model. There’s no way to really tell without seeing how the model is laid out in the game format. Either way the difference in how efficiently the hardware handles these two versions of the model should be very minute and not worth sacrificing the visual quality of the higher res version of the fridge.
