GT3 vs GT2 EPDM - Belt Recommendation
Unfortunately, I've only been able to do 12mm AWD testing and it would make more sense to do this in a more belt-bottlenecked scenario. Similarly to CPU gaming performance testing, one should run the tests with the best available GPU and in 1080p with low graphics settings. Which in this case would be 6mm belts with the most overbuilt frame and motion system. On the other hand, the narrower the belt, the more its performance will be impacted by the loss or gain of a single string of GF in its core.
Regardless of the accuracy of my numbers, it's been painfully clear that we shouldn't use GT3 for our application. While a part of the GT3 range with bigger pitches has likely been fully upgraded, it seems that 2MGT still uses the same tension members as before. The increased load ratings for GT3 2MGT likely come from the new base material and more resistant belt teeth. So in a tension member-limited scenario, the belt will have the same stiffness (~IS perf) as GT2.
This is where the problem lies with GT3, it has a significantly softer base material compared to its predecessors, which means it'll need a lot more elongation and tension to become tension member-limited. This will show up in more unnecessary mechanical wear, longer needed tensioner range, and increased rolling resistance on the idlers (lower max. accels), only to match GT2's IS performance.
I'll try to help visualize it with a graph. The slope of the tangent line is proportional to the IS perf we can get at a given tension, but GT3 needs more elongation and tension to get to the same steepness.
The next pic is the core of a 12mm EPDM belt next to a 12mm GT3:
This experimentation I've done is in no way scientific enough to declare that these belts are identical in IS perf at their respective spec tensions, but it's more than enough to show that there's no reason for us to use GT3.
If you want the better temp resistance, don't want goo buildup on pulleys, and can afford it use EPDM, otherwise just stick to normal GT2.