If you want an official response, call EVGA on your phone. This is a user-to-user public forum where forum topics do not necessarily catch the eye of EVGA employees.

I should also point out:
1) The voltage reported in GPU-Z is not necessarily accurate. You should measure the actual voltage of the GPU core using a multimeter. It may be that the card, the drivers, or GPU-Z stop reporting voltage changes above a certain point.
2) All cards are voltage limited at some point. When you hit that limit, there always is a reason why the manufacturer doesn't allow you to go higher. Every component has a specific safe operating range.

I heard that about allegedly software not being able to read voltages above a certain value. Not sure whether this is actually the case. I guess a primitive way to test more voltage would be set fan to a static RPM and then increase voltage and look whether the card gets hotter? My #*%$!* SC ACX2.0 (which is in RMA now and hopefully gets replaced) definitely was locked at 1.212C. From that point of view you're already lucky your card does 1.237, whether the 3975 is locked I don't know either.
 
As for the Epidas..sorry I also don't have exact information but common sense would say they would be rather at the same speed as Hynex or Samsung, that would be a theoretical 8ghz...and real more 7.6-7.7. However, know that the speed doesn't mean too much since it's possible that memory might have different timings, depending what type is used. "Some people" on forums say Elpida "OCs slightly better than Hynix"...for what it's worth. So or so, I wouldn't go over 7.6/7.7 on ANY type of memory simply to avoid the chips performing internal CRC correction. (No point in trying to achieve 8300 if it results in CRC errors and performance loss..then rather a notch below the max. rated speed).
 
Edit: I am surprised they must also have improved the fans itself considerably, not only the heatsink and VRMs. (Or possibly use slower fans). Because on the old models, anything over 40% is really unbearable. What RPM is the fan going at 75%?

Im going to try your primitive method.

thanks all. I dont know if reistalling AB did the trick on the new Nvidia driver did. Whether or not its accurate or 25-50 lower than actual, its reading up to max my BIOS mod allowed for. I've got my OC sorted for now.
Here is the law of diminishing returns in action
Stock 1366/7010 115 power limit 1.212v = 59.3 FPS Valley extreme 8XAA 1920x
         1461/7424 115                  1.236v = 63.1
         1556/7762 150                  1.261   = 66.5
Anything higher on core or Vram crashes, And I dont want play with more voltage for now.
I may just stick with  1461/7424 and try some hours of Cyrsis 3. As I have not tested any OC for very long under CRY3.
 

Some Old Dell CRT 19 in that does 1920 1440 at 75 hz. Im getting good FPS in Cry 3 Ultra Maxed, but not constant 60 and for sure not 75 constant. Besides that Im gearing up for Witcher 3, Dont know what I'll get in Far Cry 4. My Cpu holds my minimums back somewhat, only a i5 760 @ 3.911. The card is on Air case is open for now. But Case has plenty of fans. 2 front one huge top one high RMP rear exhuuast, one side fan.
 
Case Antec 902 MOBO EVGA P55 SLI A74 Bios
PSU Antec BP550Plus VIDEO EVGA 970 SSC 2.0 1542/7694 125% PWR +12mV 1.237VDDC
CPU i5 760 @3.9GHZ 1.231vcore 1.2vtt CPU Cooler Cooler Master Hyper N520
Ram Corsair CMX8GX3M2A1333C9 2x4GB @ 1490 MHz, 8-8-8-20 CR 2T 1.5v
Monitor Some old Dell 19 in does 1900 x 1440 at 60Ghz though
SSD Crucial BX100 250GB
Optical LG DVD+R
OS Win 7 64 Home premium Valley ExtremeHD 1920x1080 8xAA 2750

The thing u may not realize is that a CRTS have a different aspect ratio than flat panel LCDS etc. They are closer to a square. No money yet for LCD but its coming. I bench at the standard everyone in the "new world" with shiny new monitors does. 1920 1080 is not aspect ratio thats displays properly on CRTs. Niether is 1920 1200. I dont have a problem with playing at lower res. I just want to get damn dollars worth and get everything i reasonably can get. I know Im not going to set any records, LOL.

U trolling me? I never said I wasnt getting my monies worth, the fact that the card does anything at all over stock is just bonus. I KNOW THIS. I use CRTS because they are free and I have access to them, and when the image quality is still looking very good, with no chance for a dead pixel or bleeding, Its a no brainer. Until a few months ago I had a 23 in CRT that looked very good and did resolutions that only cards that costs a kidney could run at MAX. Most people that look at benchmarks dont give a crap about CRT resolutions right? thats why i bench at LCD res.

ok ty see ya
 

Makes sense, also, what I suspected. (SSC and FTW have a different, programmable voltage controller). They are also right with their assessment of memory brands, this "Samsung is better" is nothing but a myth. My last GTX 660 had Hynix RAM, some say Hynix "overclocks better" and I got it up to +475 in Afterburner, so almost 7000 from the the 6000 default. And I see THIS Samsung Ram on the 970 also handles 450-460 at least, so it's on par with Samsung.
 
As for voltages again, I have seen BIOS specifying 1.600V, some even 1.800V as upper maximum in the voltage table, so it could be assumed that some voltage controllers (in theory) could provide that much, given you have a tool that supports the particular controller and given you "unlocked" the BIOS that you can provide that much voltage. (Not that 1.6V or 1.8V would be anywhere near a sane voltage :) I don't think I personally would go anywhere above 1.25V, that is if I had a card that allows changing voltage and given than temps at whatever voltage would still be reasonable..AND given a card's layout can actually handle a wide open TDP/power draw...which of course would also have to be modified in the BIOS first. You sure don't want to draw lots of "illegally high" voltage from some puny card with only two 2x 6pin PCIE connectors, there is simply a limit what your hardware (voltage regulators), board, PSU, Cooler etc. can safely handle....and SOME of the cards are already very much at tolerance, at least from what I see.

1.275 with power limit 150% crashes the driver under heavy load. So I think you are right again.  1.26 seems to be very close to my limit. I'll stick with 1.26.