Installing the older fixed rate firmware will not let you adjust voltage or fan speed.

Just an update, I have two S9s running at 706m, hashing at 15.1 Th/s for a full week and no problems. HW errors at .0002%, low temps.

My thoughts exactly. Would love a few more machines, but not at that price.

I originally, accidentally did the firmware reversion to the wrong miner, I got my open windows mixed up and loaded the fixed frequency firmware to a perfectly good 14TH/s miner.
I freaked out at first because that wasn't the miner I was willing to risk.
But I actually didn't have any issues reverting that miner back to the autotune firmware.
It does appear to be hashing at a lower frequency then before but still maintaining 14+TH/s.
Orgininally the good miner was hashing between 649M and 661M, but with the good miner back on autotune it's at about 649M and 656M, but as I said, still maintaining 14+ TH/s.

The sickly miner still appears to be running good maintaining 14.2TH/s with the fixed frequency firmware loaded running at 662M.

It sounds like the firmware swap gave you a definite improvement to your under performing miner.
That's awesome. Glad to hear.

Hopefully some more people will be able to replicate the results.

I've looked for this information all over, and I get the same response, "It can't be done". Well obviously I can't leave well enough alone.

So here it is...

This worked for me, it may work for you.
You can do this if you want, but if you end up bricking your machine, don't come crying to me.

For those that don't already know, the first batches of S9s came with the ability to change the frequency of the machine, thus increasing the hashing speed, however doing so would void your warranty.
That is still the case today, overclocking your Antminer will void your warranty.

OVERCLOCKING YOUR ANTMINER WILL VOID YOUR WARRANTY!!!

As some of you know, the Antminer S9s that have been released in the last twelve months are autotune units, and as such the advanced configurations menu is not immediately visible. You can get to the advanced configurations menu by simply placing "minerAdvanced.cgi" in the following way:

http://192.168.xx.xxx/cgi-bin/minerAdvanced.cgi  (Obviously the router assisgned IP of your unit will be your own.)

This menu really consists of only one thing:

"Search Frequency From": Along with a drop-down menu listing various frequencies.

Now if you have an autotune unit, you can do anything you want with this, and it won't change a damn thing. It'll act like it's getting ready to do something, but it actually does nothing.

So, here's the back-story.
I purchased three Antminer S9s from Bitmain to be delivered with the Nov. 2017 batch. These were 14TH/s models.
I received said miners, hooked them up, and everything was hunky-dory, until it wasn't.
All three machines worked just fine for the first couple weeks, and I do mean a "couple" weeks, as in 2 weeks.
Then one of the three miners started throwing up a bunch of hardware errors. So, I rebooted the machine, but it didn't help,
so I did a soft reset from the miner interface, it didn't help, I did a reset by sticking a paper clip in the little reset hole on the front of the miner, it didn't help either.
So, I did a hard reset by unplugging the machine, pushing the IP Call button, keeping the button depressed while I plugged the machine back into the outlet, and kept my finger on the button for another five seconds after plugging the machine back in before releasing the button.
And guess what?? It took care of my hardware errors, totally no more errors.

But now my brand new 14TH/s machine was only hashing at 13.5TH/s.

Now we already know that Bitmain sells us used machines, that's not really a secret, or that they take 13.5TH/s machines, overclock them to 14TH/s, put "autotune" firmware in them so that we can't overclock them further, and then try to act like we're getting a brand spanking new 14TH/s miner.

So, now I was being faced with a dilemma, pay to ship my Antminer back to China for repairs, and wait 2 months before they get around to sending it back to me, or send me a different unit altogether, or just accept the fact that somebody did a shitty job converting this unit to a 14TH/s unit, and just continue to mine with it, after all, is there really much difference between 13.5TH/s and 14TH/s.
So that's what I did for a couple months.

Well, obviously among other things, this coalition of big corporations such as Google, Twitter, Facebook, Mailchimp, etc., in concert with old Wall Street dinosaurs, attempting to suppress the crypto markets in an attempt to control the will of the people, it's painfully obvious that the machines that in early January were bringing in revenue in excess of $36 per machine, per day, is now only bringing me around $6.95 per machine, per day.
So, I finally said, "Fuck it!!"

So this is what I did:
My handicapped miner was only hashing at a frequency of 626M, compared to my other two 14TH/s units which were hashing at a frequency between 650 and 662.
So, I went to the Bitmain website and downloaded the firmware for the "650M fixed-frequency" S9s.
Knowing that I could potentially be bricking my miner, I was willing to take that risk.
I pulled up the miner interface and loaded the 665M fixed-frequency firmware and sat back and waited. The miner rebooted, and LOW & BEHOLD!!!

My sick miner was now working at 650M and hashing between 13.9TH/s and 14.0 TH/s!!

Fuckin Sweet!!!

So, I said, "why stop there?"

I went to the advanced configuration menu and set the "Search Frequency From" to 662.5M, hit "Save&Apply"

And guess what, my once sick miner was now working at 662.5M and steady hashing at 14.2TH/s.


So to recap:

1.) I loaded the 650M fixed frequency firmware into my autotune unit.

and

2.) I changed the frequency to 662.5

And Now

I am hashing at 14.2TH/s at 662.5M, from the original 13.5 TH/s and a frequency of only 626M.
Temps are normal and hardware errors are around .0004% which is actually an improvement from .0008%

Like I said, this worked for me, and it may work for you, or you could end up bricking your machine.

But know, you're doing so at your own risk. And I cannot guaranty success.

But it could be worth a try.
It could be the Holy Grail of
Overclocking your Autotune Antminer S9.

There is one other thing, once or twice an hour every couple hours, sometimes more. I get a red fault light on all three miners at the same time, sometimes it'll just be two miners. I thought maybe my Ethernet switch might be too cold as I have them in my garage. It's a 5 port Netgear ProSafe FS105. It states that it's operating temperature is between 32F - 104F. It's been freezing outside, but I seriously doubt that it's actually been down to 32F in my garage, especially with three miners throwing off so much heat. Nevertheless, I set up a small 250 watt heater pointed at the Ethernet switch. It sufficiently warms the switch without adding to much heat for the miners. But I'm still getting fault lights on all three miners and the kernel logs say, "Fatal Error: network connection lost!" They all eventually go back to green on their own, but this can't be normal.
  What makes me curious is that all this started when I switched from Slush to Bitcoin.com. Before switching to Bitcoin.com I never had any of these issues. Which is why I ask if any of this could be caused by the pool.

So, I have three S9s from the Novemeber batch, all three are 14Th/s. I haven't had any serious problems to speak of other than in the last 10 hours one of the boards has racked up 121,295 HW errors or 0.1076%. The percentage has actually dropped, it was at 0.12 The board continues to hash at 4,733.24 GH/s and everything seems normal with the pool I'm mining in, Bitcoin.com.
The temps for all three boards are 74, 71, 67. The board with the errors is the one at 67.
On the miner status page there are no "x" in the ASICS status section. Can something within the pool do this, or would this be a board problem.