Overclock AntMiner S9 14th/s

[MinerMEDIC]

Legit - i tried it... 20% on L3+

Stock          384M @ 500 mh/s
Overclocked 462M @ 600 mh/s

earnings reflect that too....

and well to the person asking what if i increase output by 20% and decrease life by 50%... i would say they become almost useless with difficulty change in the latter 50% of their life at which point there is more efficient and powerful machinery.... please tell me if i am wrong - i AM still new to this... and did that guy ever overclock his machine?

If you brick it you lose all the resale value. Go check out what S7s are going for on Ebay. I made those % numbers up, just as a way to illustrate that overclocking, to me, appears to be a short term goal because you sacrifice the life of the equipment and most likely completely void the warranty. I am sure Bitmain can tell an overclocked burned up miner easily.

Funny thing is nonfunctional hardware still fetching an impressive amount.

[Raymond_B]

Legit - i tried it... 20% on L3+

Stock          384M @ 500 mh/s
Overclocked 462M @ 600 mh/s

earnings reflect that too....

and well to the person asking what if i increase output by 20% and decrease life by 50%... i would say they become almost useless with difficulty change in the latter 50% of their life at which point there is more efficient and powerful machinery.... please tell me if i am wrong - i AM still new to this... and did that guy ever overclock his machine?

If you brick it you lose all the resale value. Go check out what S7s are going for on Ebay. I made those % numbers up, just as a way to illustrate that overclocking, to me, appears to be a short term goal because you sacrifice the life of the equipment and most likely completely void the warranty. I am sure Bitmain can tell an overclocked burned up miner easily.

Funny thing is nonfunctional hardware still fetching an impressive amount.

Solid point, people are selling miners for parts and asking/getting dumb prices!

[shefkk]

i have a bunch of S9s - if i can overclock those too, it would be amazing...
saw old vids saying to set frequency to 800, any feedback? 

oh and in regards to warranty - with my shipping costs and wait times, i would not be able to utilize it anyway.... but thanks for the input
had a bad experience?  can i learn from your experience?  i mean it seems to be working fine... just making sure temps and power are good...

[shefkk]

i tried it

[HoleShot]

So?
What happened???

I do have a few suggestions.

First, only use one hash board. Maybe even use a has board with one or two ASIC strings dead. So one that's barely working so there's much less at risk.

Install old firmware to get control of the clock.

Reduce intake air to <15C. Every 10C temp reduction should double lifespan.

Start at 600M and increment clock the smallest increments possible with 1 hour run at each setting. Document all parameters. Continue to increase clock rate until excessive HW failures, >1000 within the hour occur, or ASIC strings dies, whichever comes first. 

Report findings here to give others an opportunity to learn from the data and create a new tactic. Build knowledge base to allow aggressive miner types to maximize benefit.

 

[HoleShot]

http://192.168.x.x/cgi-bin/minerAdvanced.cgi (change ip address to your miner's actual address)

There's the direct call to the Advanced tab. That's really cool as you don't have to use old firmware.

I went there and it offers the starting clock scan frequency as the only menu item. Mine was set at 550MHz but my hash boards are running at 665. I guess it started at 550 then worked up  to 665. Each hash board looks to be independently controlled. I have one much older hash board in this miner which is running at 550 while the new boards are at 665.xx

When I selected 575 as the starting clock speed and hit Save,  a warning popped up saying;
It will take about 3 hours on search frequency before mining! Are you sure?

hmmm. I'll set a miner up with a single hash board to play with.

[shefkk]

ok well i somewhat failed.... nothing happened to the miner.... but.... nothing was faster than auto frequency (speaking of -- to the comment above me, you are on autofrequency firmware.)

like you suggested i had to go back to older firmware to control frequency; however frequencies i chose resulted in lower hashrates... tried 800, and 812 i think was the next but registered as 850 on status page....

do you think there is a frequency that could enhance it? i realized 650 even slows it down.  Does it take time to adjust to new frequency?

[Sandal_Hat]

This is a very bad idea. Reliability and safety is more important.

[HoleShot]

Hmmm. not sure. The "ideal" rate is just a math function with the "RT" result is measured. If the RT is very near the ideal then all is working well.

As an example I have an old and new board in one miner with auto freq.

Freq.     GH/s(ideal)  GH/s(RT)
550.00   3950.10      3974.35
666.28   4785.26      4709.43   

The auto freq board sets the older board at 550 by default. It does not try to optimize the clock speed on this older board. I used to run this board at 625 on the older firmware.

When you increase clock speed you should see a proportional increase in ideal hash rate. If the RT rate is far from the ideal rate you can see the clock rate is causing some problems.

I'm not sure the clock setting on the new firmware sets the minimum clock speed as it says the starting clock speed. The firmware may reduce the clock speed from this starting point to find the optimum freq. I'm thinking this setting may get the miner started sooner upon restart if I program a starting freq very near the found optimum frequency. I may give it a try on one of the miners with 3 new boards in it.