So I'm doing some voltage fudging today, and mess around with input volts. Some people were trying low input voltage and getting rather low. Well I went to 1.675 (Didnt try any lower) and with 4.5GHz core/cache, 1.3v core, 1.2875 cache, there were some differences between 1.675 and 1.95 input.
Firstly, not a difference, both were perfectly stable on prime95 (I really wanted to push power consumption to try to get input voltage to be the fault).
Low input voltage vs high input voltage causes a much higher VRM temperature. With a 60mm 4k rpm fan blowing directly on VRM, I got 79C peak with 1.675v input, and 66C peak with 1.95v input. This is likely due to higher current being pushed through the chips to make up for the lower voltage.
The FIVR does not seem to care between low and high. Low vs high showed no discernible difference in power consumption at the wall, and no discernable difference in CPU temperatures (Using a strong liquid cooling setup, Coolant was 31C steady).
The only other difference i saw was estimated CPU Package Power, where estimation was 11.44% higher with low input volts vs high input volts (363.548W vs 326.223W respectively)
This leads me to think that while FIVR might not care what voltage you're feeding it, you're better off giving it a higher voltage than lower for the sake of keeping your motherboard's VRM temperature in check, and keeping input current per CPU pin lower.
Also unrelated to the above, but I've just tried BIOS 1902 for the RVE, and it fixed my 3000MHz RAM divider for 100MHz BCLK @ Room temperature. I could not for the life of me get it stable again since it was subzero (BIOS 1802?).
It also seems to be pulling more power @ the same volts, which is why some people are claiming they need less volts now (maybe internally pushing more volts than it tells you? who knows)
Edited by ssateneth - 2/27/16 at 3:29pm