Hi all
 
I'm confused. Just been doing some benchmarking with my 3080 XC3 Ultra. With a modest over clock of 64 on the Core and 200 on the memory clocks, I get a score of 11285, in order to see why I was a lot lower than others I guessed that amongst other things was temp as I was averaging 61c. So I wacked up the fans to 100% and ran the test again. This time I only scored 10840 but a temp of 57c. I thought this was odd as I figured it would be better but then I thought I was power limited as the fans at 100% would draw more juice away from the GPU. So I upped the power to its max of 108 and ran again. This time a score of 10806??? Went back to normal fans and 100 power limit and back to a score of 11270??
 
I then left the fans at normal and put the power limit back up figuring that normal fans, more power would give higher score but still 10800
 
Can anyone explain this as I'm confused?

Yes sure
 
Run one - Fan curve - extreme, modest OC - Score is 11285 average GPU 1872
Run two - !00% fan, modest OC - Score is 10840 average GPU 1896
Run Three - 100% fan, modest OC, max power (108) - score is 10806 average gpu 1880
Run four - fan curve Extreme, increased OC of 94 on the clock, mem still at 200, max power (108) - score is 10884 average gpu 1903
run five fan curve extreme, increased OC of 94 on the clock and 300 on the memory, max power (108) - score is 11903 average gpu 1901
 
then I ran a run on default settings: Score of 9971 average gpu 1813
 
So it seems the lower OC, with lower power limit gives the best score??

I'm not a Port Royal or OC expert by any means.  I've only run port royal ~30 times on my 3080.  But there are two things I have noticed as a trend, first is the score (for me) seems to scale with average GPU clock, and the second is occasionally I'll run a 12,XXX then my next run will be 2-300pts lower with no changes, no temp change or anything.  I think is that is indicative of my OC being right on the edge.  I pull back a bit on the GPU clock and I won't hit that max #, but I'll be really close and can pretty much ensure repeatability.
 
The other thing I have seen is that if I am experiencing that wide swing in scores I can kick off a Time Spy run and it will fail every time.  If I have consistent Port Royal scores from 2-3 back to back runs then I am pretty safe with Time Spy completing.  
 
Lastly, I have a buddy who advised me to ditch EVGA X1 and try MSI Afterburner.  I am getting far better and more consistent runs with my afterburner setup utilizing their voltage curve mapper as opposed to just a raw clock offset (which is all X1 has).  

I get about the same score, 10770 with a XC3 black 3080. I can't seem to use afterburner due to the fan not being controlled properly on that one.