1. Reference cards are gimped in that they run at or above the thermal throttling point at stock settings. The blower style cooling system is simply not powerful enough to adequately cool the card.
2. In addition to cooler design the reference cards are further gimped from the use of the reference PCB. AIB cards will include various "improvements" such as beefier VRMs, thermal pads, improved chokes, etc.
3. All designs exhaust some of the air outside the case .... blower style coolers exhaust more of it because the shroud is designed to do so. Because the air has to be squeezed out that little tiny hole in the back, significant backpressure is developed and air does squeak out thru shroud openings. In addition, the card itself radiates heat in all directions. Think of the card like a radiator ... a water cooling radiator disperses about 60% of the latent heat in the system thru the rad fins and outside the case.... the other 40% radiates from he surface of the radiator shroud, tubing, fittings and the components themselves. The shroud, PCB / Backplate all radiate heat out into the case.
Another thing to consider ... Look on newegg at CLC type coolers for CPUs and we see models with 1 to 3 fans. So how is it then that a CPU cooler needs 2 or 3 fans and yet a GPU CLC type cooler only has 1 ? Shouldn't a cooler for a 1080 Ti that can burn 350 watts require more radiator and fans than a 90 watt CPU (130 OCd) ? The reason it is able to maintain lower temps w/ just 1 fan is that so much heat is radiated from the card.
4. I certainly understand the mindset that says it's bad thing to have the heat inside the case .... but, assuming for a moment that it actually stays there, what component are we worried about getting 2-3C hotter ? I just ran Furmark (100% load on twin GPUs) for 20 minutes .... start to finish:
CPU went from 37C to 39C
MoBo went from 40C to 40C
GPU went from 41 to 58C
SSD1 went from 35 to 36C
SSD2 went from 38 to 38C
SSHD 1 went from 38 to 38C
SSHD 2 went from 39 to 39C
That is with water cooled CPU, MoBo and twin GFX cards ...
all of which are cooled with radiators equipped with fans blowing all that heat into the case.
The reason it has no impact on any component temps is because the case, the entire case volume is turned over more than twice every second. So as long as you provide case ventilation, the interior case air will be replaced 1 - 2 times a second. The heat doesn't stay in the box because it is gone before it would have any chance to affect anything. Ya can argue with the logic, but the argument ends when ya look at the numbers above. There is simply no impact and in this box, we're talking over 700 watts of heat being generated.
If you use a generic case with 1 rear fan, an interior case temp rise between a blower style might be measurable. But with a big GFX card, the standard 3 intakes / 2 exhaust will show no impact. OTOH, negative impact (throttling) on the blower style GFX card is immediately obvious.
GTX 980 Reference
MSI 980 Gaming X
That temp hitting the throttling point limits the performance of the reference card which is why the reference card hits < 150 in the OC test and the MS breaks 160 fps.