I'm glad people found my post useful. I'm a huge geek and I hope to use my powers for good, so to speak. The alternative is not using them for evil but rather using them to bore dinner companions to death.
Fixie82, the only valid answer to your question is "it depends, but generally, deeper is better."
I've uploaded a screenshot from Flo Cycling's aero page:
Capture.JPG [ 126.11 KiB | Viewed 235 times ]
It shows that drag at zero yaw is pretty much the same for any of Flo's aero wheels, but also that the differences in drag (unsurprisingly) increase with yaw angle. When I sat down to write this reply, I expected that even at amateur sprint speeds, you're going fast enough that yaw is usually pretty small. But then I did the vector math, and yaw matters more than I thought. Even if you're sprinting at 40 MPH a 5 mph cross/headwind at 45 degrees to you creates an apparent wind angle (yaw angle) of 5 degrees. If you're going 37 MPH (as in my previous example), the yaw angle is about 6 degrees. When you start the sprint at about 30 mph, the same crosswind gives you a yaw angle of 7.5 degrees. If you look at the Flo chart (sorry!) you'll see that the deeper wheels can pay off, even in a sprint.
At 7.5 degrees, the 90mm wheel has 33% less drag than the 45mm wheel. That's only 18 grams. If you double that for two wheels, you're looking at 36 grams difference at 30 mph. I've seen heuristics that suggest 50 grams of drag at 30 mph is worth 5 watts, while Flo claims those same numbers work out to 6.5 watts. Depending on who you believe, the 90mm wheels are worth between 3.5 and 4.5 watts over the 45mm wheels at the start of the sprint with a nontrivial crosswind. In the context of a 1100-watt effort, that's not much. As long as you have reasonably deep, reasonably aero wheels, you've got most of the benefit under most circumstances.
Myself, I'm saving my pennies for a disc-braked aero bike, so it doesn't make any sense to get a new set of rim-braked aero wheels now. I've got at least a year to save...I'm waiting for Trek to release a disc-braked Madone, though I'm also intrigued by the 3T Strada. No matter what I get, I'll probably end up using Light Bicycle's 65mm U-shaped rim in the back and their 55mm U-shaped rim in the front. If I had more cash, I'd probably get a set of ENVE 5.6s or 7.8s. The difference in weight for the disc versions is about 300 grams, which (as we've seen) doesn't affect acceleration very much.
I'm curious about what kind of difference aero wheels plus an aero frame like the Madone would make. Flo has a useful chart here
. It shows the effect on a rider's velocity for a 100-gram reduction in drag. It's a gross oversimplification, but it's good enough for our purposes. Trek claims in their Madone white paper
that between zero and five degrees of yaw at 30 mph, the top-level Madone (sporting deep-section Bontrager wheels, I assume) has about 300 fewer grams of drag than a "standard road bike." I'd take this to be maybe a moderately oversized round-tubed frame with 28-spoke not-especially-aero wheels and a non-aero handlebar.
If we assume that Trek rounded up* and we in turn round down aggressively, let's say that the deep-rimmed Madone you're on has ~175 grams less drag at 30 MPH than the round-tubed Crumpton with Ksyriums that the guy next to you is riding. If 175 grams less drag gets you 1.33 times the 100-gram speed differential Flo lists--remember, drag and the power required to overcome it increase nonlinearly with speed--then you're looking at an increase of 0.57 MPH over your competitor when you start sprinting at 30 MPH. That works out to 0.254 meters per second. The sprint I described in my first post took 13.5 seconds to cover 200 meters. If you're inching ahead at 0.254 m/s over the guy next to you, after 13.5 seconds you're 3.43 meters ahead of the other guy. That's over 11 feet, or two bike lengths.
Of course, the Madone is on par with the best aero road bikes, and the bike I put your competitor on is a little out of date. Realistically, you're racing against people with deep-section wheels and quasi-aero frames. Your advantage over those people (assuming they're your height and weight and that they make exactly the same power you do) is going to be much smaller...somewhere closer to 1-3 feet. But I've lost plenty of races by a foot and a half, and I've won a few by that much or less.
And that's just the sprint. That doesn't count the matches you didn't burn earlier in the race because you only had to make 500 watts to jump out of each corner instead of the 520-530 watts your competitor had to make.
To answer your question more directly, Fixie82, most aero wheels are similar at zero degrees of yaw. Small yaw angles between 5-10 degrees are pretty common, even at high speed (30-40 MPH). In zero-degree-yaw conditions, a 65-90mm rim doesn't really hurt your acceleration measurably. In a sprint involving a crosswind, they'll help measurably. Objectively, you want to go as deep as you can in the back and as deep as you can handle in a gusty crosswind for the front.
All that said, my current bike is about 13.75 pounds (6.25 kilos) in race trim, and as much as I'd love to own a ~7.4 kilo Madone, I'll still miss hefting my weight weenie bike on sleety winter days.
* I've done FEA and CFD for most of my engineering career. That white paper is the best, clearest, most convincing account of properly validated FEA and CFD models I have ever seen. Trek's simulation group is being very, very meticulous. I saw very little that was either sloppy or too perfect. I have no insight as to how accurately the white paper conveys the results they actually got, but wow...all the i's are dotted, the t's are crossed...it's solid. If they're fabricating data, they're putting more work into the fabrication than it would take to get very good real data.