serial just looked cleaner in the case.  I have a flow meter on the line and can boost output in a GUI so it didn't really matter.  I monitored temps on each card and found the difference was negligible.   In the future,   I'm just going to buy one heavy hitting card and run one card.   Things have gotten so fast, One GPU will do it.    At the time I purchased those, it would handle any game in sli at stock settings, so I didn't need to overclock them.  So heat really wasn't a factor. 

GGTV-Jon
Along with the pressure versus flow is in series (generally the bottom card is fist just because of tubing routing) the first card gets the cooler water then the second card gets cooled. Where parallel both cards get the same fresh intake flow. How much of a difference depends on what the work loads are and the water blocks I guess.
 
In reality, yes once the loop gets up to temperature thing should normalize. How much of a temp difference between the cards and weather or not it is affecting things is subjective and in the end it is up to you weather or not it matters.
 
Personally parallel makes more sense - it is up to you to decide

I blame ^ for changing And I will never go back to serial.
I have mine setup this way
           1                 2 

          3                   4
Temp Being this Coolant Temperature Sensor Plug, 10K Ohm
(1) Front Left In (2) Top Right Temp shows 31.5°C (2)
(3) Front Left Temp shows 28.9°C (4) Bottom Front Right Out.

GGTV-Jon

mattman657
I'm less considered about temps as I have 2 x 360mm and a 480mm UT60s. Rad capacity isn't a problem. Also running the EK Dual D5 pump, so flow isn't really. Based on how the two factors, that may influence parallel vs serial, aren't really an issue for me, I'm tempted to go serial if I do end up going SLI.

Issue is that say it is only a 1-2c difference between the 2 cards, the hotter one just might be tripping the threshold step for your GPU clocks

I guess the question is, that 1 or 2 degree difference, is it even present once your loop equalizes? I personally have never put GPUs on water so this is foreign territory for me.

Total span of vdroop settings:
 
Observed Vcore was slightly larger (~0.005 volts) at idle than what I put into the bios with negative percentages.   Vcore was slightly less (~0.002 volts) than what I put in the bios with Positive percentages. 
 
The voltage gains are these: 

-25% added 0.005 to vcore from what I put into the bios at idle
         added 0.003 to vcore under load.   Minor increase. 
-50% added 0.005 to 0.01 to vcore from what I put into the bios at idle
         added 0.020 to vcore under load.  Small increase
-75% added 0.01 to 0.013 to vcore from what I put into the bios at idle
         added 0.018 to 0.042 under load   Moderate increase
 
the voltage drops are these:
 
+90%    0.070 volts
+75%    0.068 Volts
+50%    0.048 Volts
+25%    0.034 Volts
Default   0.021 Volts
Auto       0.016 volts 
 
test performed at 5.1ghz.    C states and virtulization disabled. Windows set to max performance.    voltages read in Eleet and cpus stress was under intel extreme tuning utility.   I am sure you can try this with different CPU stressing programs.  I used the intel extreme tuning utility because the engineer of the board indirectly recommended it.