Thanks for that, what cards did you get out of interest?
Are they able to hold that 2100mhz or if not what does it drop to over time? After reading the below I'm thinking custom cards are only good for the looks and being cheaper, I'm not bothered about the price.
The information we have so far;
Reference boards with the FE bios is limited to 1.061-1.080v. The max power draw is roughly 210W. ~20W of that to the memory when at 5500mHz. The remaining 190W limits the core to about 2020-2050mHz on air. Higher requires a higher power limit than the card can provide. The FE blower fan is able to cool this OC, but requires a custom fan profile set higher than the stock config.
AIB air cooling solutions using the reference board results in the same performance. ~2050mHz max with a 5500mHz memory. Cooling is better, card runs a little cooler and quieter, but no additional performance is gained due to the power limitations of the reference board.
ASUS STRIX cards with extra power have been VBIOS modded to unlock voltage. This is limited to 1.25v. At 1.25v with 300W available and an unlocked VBIOS, the people are hitting a maximum of ~2100-2110mHz core and 5650mHz memory before instability. Core is not responding to move voltage very much. Heat also isn't increasing too much. Cards at those clocks are drawing about ~230W total. The STRIX cooler is easily able to keep up, but there is no more voltage available to use the remaining thermal headroom to try for more clock.
Gamersnexus frankensteined together a hybrid FE 1080 reference board, keeping the core below 42C under load. They did nothing to VBIOS and their card was stuck clear down at 1.061v. By downclocking the memory to 4500mHz to try to free up as much wattage as possible below the reference boards low limit, they manages a psuedo-stable 2204mHz clock. It passed Furmark, but crashed in games. They did manage a stable 2164mHz. This is notably above anything people have managed at 1.25v on custom boards without the power limit. The core appeared to be current limited and NOT voltage limited, due to the improvement in stability of the core at the same voltage when downclocking the memory to save wattage. This makes no sense when combined with the STRIX testing with unlocked vcore and 'unlimited' power draw achieving significantly less performance. Lower core temperature may conduce far more core stability than we've gotten in the past. With only a single water cooled card tested in this manner, there is not enough information to be sure. They may have had a golden chip.