My miner has stayed within the temperature range of 74 degrees to 78 degrees for the hottest Chip. When it gets to 78 the fans kick into overdrive and put it down to 74. If I keep the Antminer S9 in a cooler place, the temps drop a couple of degrees and the fans do not run in overdrive and make much less noise. I have not seen a difference in hashrate on the S9 Status Screen based on room temperature so far.
The hashrate on the Status screen has been consistently around 13.50 TH/s (Ideal) and 13.68 TH/s (RT). Not sure what RT means.
Here is my AntMiner S9 Status Screen: https://imgur.com/a/UdaSg
Now, on the pool website, for worker shifts, my hashrate has varied wildly and randomly between 11.15 TH/s and 14.81 TH/s, but the average is around 13.68 TH/s.
Since shares themselves are random - though they should average about 18 per minute - they follow exactly the same distribution model as blocks.
Meaning, that sometimes they'll be faster than the average 3.333 seconds and sometime they'll be slower.
The pool only sees shares, and calculates your hash rate per shift, based on the value of those shares and the length of the shift.
So e.g. if you submit S=2300 shares, valued at D=4098 each, during a M=50 minute shift, your hash rate would be S x D * 2^32 / (M * 60)
or 2300 x 4096 x 2^32 / (50 * 60) = 1.349 x 10^13 or about 13.49 THs
Please note that I picked those numbers to make the answer look similar
Now what really happens is that although there's expected to be around 2300 shares at 13.5THs in a 50 minute shift, there's actually, randomly, more or less than 2300 shares.
That's the share variance that makes the shift graph go up and down.
But the important point is, the shift graph shows you exactly what each shift is worth in terms of Diff - which is what you get rewarded for.
It's not some averaged out value that isn't what you will be rewarded, it's the actual Diff value, and it's expected to vary a bit.
At the moment there's a lot of shifts in the 5Nd payout range, so that variance effect on the total reward is extremely small, but even when we were larger and finding a few blocks a day, that variance is still very small since at e.g. 3 blocks a day, the 5Nd is about 48 shifts.
Thanks for the in depth analysis! I have also noticed that the shift time is not always 50 minutes, I have seen it vary up to 10%. Then if a block is found, it looks like a shift terminates early.