I've taken some calibrated DC draw and other power measurements on my one S9. I've taken some care with these and everything should be accurate to at least the precision I give.

Short version of results:

S9 DC Draw at 12.1V  == 1.89W/MHz + 57W

The 1.89W/MHz holds across at least 450 to 668.75 on this unit; the 57W is the controller and fans at 100%.

At 650 w/ 100% fan I see 106.67A @ 12.051V for 1285.5W.
50% fan reduces this to 1275W, hitting the S9 specification.

MHz |	DC W  |	AC W  |	PSU Eff.  |	W @ Wall w/ Fan  |	J/GH @ Wall
450	855	968	88.3	1033	0.107
550	1040	1175	88.5	1240	0.105
600	1133	1281	88.4	1346	0.104
625	1179	1333	88.4	1398	0.104
650	1229	1391	88.4	1456	0.104
668.75	1265	1433	88.3	1498	0.104

PSU is a DPS-2000BB being fed 208V; voltage at PSU ranged from 206.8 to 208.7 during these tests.
All measurements taken simultaneously - Dedicated sampling for I & V on both AC & DC
+12V measured at the S9 board inputs ranged from 12.051 to 12.146, depending on load.
These DC Power draw measurements below _do not_ include loses in the 12V cables - these are dependent on your setup but may be around 10W; actual PSU efficiency is accordingly around 1% higher.