Demonstrate your conjecture. Demonstrate how it works in the application you have described. I'll wait, because nothing thus far has done so. In order to do that, you'll have to find data that I'm not sure exists. Namely, data where all other items are held constant and testing is done in conditions comparable to what a CPU cooler would see. No research paper, no variation in material, and no variation in design because a 1:1 is not produced for consumer products. I know I'm asking for the impossible, but you bear the burden of proof when you claim that dimples automatically make a cooler better. I have the much easier job of poking holes into the argument.
Mathematically, a turbulent flow exposes more area to the transfer material (and the dimples increase surface area too). Heat transfer also occurs by both conduction and convective influences, effectively boosting q. In theory world, this means a maximum of 30-40% better q values because the A value is increased. I say theory world, because the paper in question did not focus itself on practical applications demonstrable in a CPU cooler. This is the issue with drawing a conclusion based on poor arguments, which derive themselves from a poorly interpreted appeal to authority.
Now in the real world (read: messy and prone to us humans borking things pretty badly), let's consider doing this. You are a brisk business, and sell about 100,000 units of a cooler per year. Said cooler requires only 30 fins per unit. That means an additional 3,000,000 processes per year.
I'll give you a press which can produce 2 parts per stamp, because any more than that and the labor of loading and unloading starts to add up. You need 1,500,000 cycles. The average laborer does 2000 hours of work per year, which means that if this takes an entire shift all year round then you've got to have 750 cycles per hour, or 12.5 per minute, or a cycle every 4.8 seconds. Pay said laborer the equivalent of $30,000 per year. Now, purchase the press. You're looking at an initial investment of maybe 250,000 dollars. Consumable materials, including the dies for the stamping, are another 30,000. Let's just put maintenance due to wear at 45,000 per year (swag, but based upon oil, minor fixes, die work, and other factors). You've got $355,000 invested into adding dimples to your 100,000 units sales. A net cost of $3.55 per part, without even factoring in inefficiencies and down time. You're looking at an increase in cost of about $5.00 per part once the increase in infrastructure is actually factored in.
Now, build in profit. You produce the parts, and sell them to a distributor with a margin to make this work profitable. The distributor adds their margins, before selling it to a retailer. That $5.00 cost was increased to $6.00 by you, $12 by the distributor, with $24.00 by the time you buy it. If you have a cooler that costs $30 to buy, with a $24 upcharge because of an additional process, you have functionally 200% costs for a 140% performance.
Now, let's look at your one example. The cooler is at most 1 degree C cooler than its competition. The noise level is incomparable, because the fans aren't certified to be uniform. Benefit of the doubt, a 12% increase in relative cooling. How much is the increase due to change in effective area, convection, and changes in material thickness (s)? I can't determine any way to tell, but if you see something I'm missing I'd gladly acquiesce to ignorance in order to understand your perspective.
Circling back, walk me through this one more time. Exactly why would you suggest we add dimples? In experimental conditions, with fine controls that are all but impossible in regular manufacturing, they demonstrated a 30-40% increase in q with a higher pressure drop. In practice, you demonstrate negligible performance gains with a test that is eight years old.
By this same logic, we should really use silver instead of copper, because that extra 7% k value will make a difference. It's odd, because you hit the nail on the head earlier, aluminum is used in fins not because of its performance but because it is less costly and easier to manufacturer. Yet in the case of dimples, you've reversed your opinion. A minor, demonstrated by a single degree in ideal conditions, performance difference has a large associated cost.
As a note, the cooler in question was Japanese only. A country notorious for spending vast sums of money on perfection, rather than having a reasonably priced 80% solution. I can't find any pricing information, but I'd conjecture that was the case here. Again, if you can prove me wrong I'd gladly acquiesce to ignorance.
-edit-
As to the math again, welcome to heavy physics. Namely, the world where scientist acquiesce to their models not being an accurate prediction of the world.
Read your own section please, because while rather wordy, it spells things out....clearly if you understand the background. To the layman, this is all insanity:
"In an effort to disrupt the boundary layer in continuous fins, perforated fins have also been studied where a pattern of spaced holes are formed in the fin material before the fin is folded into a U-shaped flow channel (Webb and Kim [2005]). The perforated fins produce little heat enhancement in the laminar flow regime and a moderate one in the turbulent regime (Webb and Kim [2005]). Fujii et al. [1988] studied a new type of perforated fin, where the heat exchanger is constructed with surfaces using enlargements and contractions forming a trapezoidal shape. Fujii et al [1988] covered Reynolds number flows less than 3000. They reported that the heat enhancement from the fin surface is due to the secondary flow induced by the suction and injection through the perforations, and due to the frequent boundary layer interruptions at each contraction part."
Note:
1) The perforated fins produce little heat enhancement in the laminar flow regime and a moderate one in the turbulent regime (Webb and Kim [2005]).
2) Fans produce functionally laminar air flow
3) Dimples generate turbulence, assuming some high velocity air
4) The cited Reynolds numbers are high. High as in not applicable to our geometries high.
5) The cited paper,
http://heattransfer.asmedigitalcollection.asme.org/article.aspx?articleid=1443782, deals with tube heat exchangers and surface roughness
To put a lot of science very briefly, it has been stated that you can increase heat transfer by turbulence. You can generate turbulence in low density and low speed fluid flows vie discontinuities. Even despite this, the influence ranges from little to moderate. Given that we aren't talking huge temperature differentials, this equates to low gains in practice.
The funniest part, none of this matters. The Reynolds number for a chord length of 15 cm, with a 100 CFM fan, whose linear velocity is 3.5 m/s, with a kinematic viscosity of 1.5111E-5 M^2/s, is 34,743. Way outside the range of the 3,000 cited.
Again, the papers are not looking at the same thing you are (Reynolds calculator here
http://airfoiltools.com/calculator/....15&MReNumForm[kvisc]=1.5111E-5&yt0=Calculate). If you could slow the flow, increase the chord length, or increase kinematic viscosity we'd be on the same page. As this stands, you're still not applying the research in a constructive manner.
While technically correct, you're spending a huge amount of effort to find the last 2% of returns. It isn't done, because it isn't rational. If you want to argue that, then please build your own superior cooler for the same price. I'll vote with my wallet, but I'm not paying 200% of the price for 140% of the performance. I will however agree that thicker fluids (read: liquids) have this make a heck of a lot of sense. If you want to build new type of cooler I'm on-board.