[GUIDE] Undervolt antminer s1 [1.19W/GH at the wall]

[SLEI]

I undervolted my S1 V1.3 from year 2013 (very thin metal frame) with 3.3k resistors. Voltage is 0,82 and I´m hashing now with 250mhz and 125GH 2% error rate in very warm room. Measurement point was different than in first page picture, I measured from top of low row of capacitors.
I was shure first that operation was succesful but patient died as miner didn't work with 193 or 200mhz

Power consumption 170VA from wall.

[klondike_bar]

I figured it out, for some reason even though I dropped each resistor to 2.8-3k with my pencil they all were somehow even lower down to 2.5k after a few days, i wonder if the lead can heat up and do something weird which would lower the resistance even further?

^that does happen. Its something that happens as the lead 'cures' and becomes less resistive as it does. Same thing hapened pencil-modding bitfury units - after a day or so from modding it would become unstable due to higher voltage.

IME: after ~6hrs the 'curing' is about half-complete, at 12hrs about 75%, at 2days you should be totally stable and not see it stray any further. Its worth keeping this in mind when modding, particularly if the unit becomes less stable after a day or so from when it was modded

[squashpile]

Undervolted two S1's last night once I got S3 online. I guess I did ok. I'm running these two on HP server power supply with breakout board. I have not checked power usage at wall yet.

Clocked @ 250 does that look about right?

S1/1 Mining BTC at P2Pool: 101.56 Gh/s  (121.79 Gh/s avg)   A: 3925  R: 85  S: 34  HW: 5980  WU: 1701.41  U: 3.58  LW: 2889681
S1/2 Mining BTC at P2Pool: 137.68 Gh/s  (127.79 Gh/s avg)   A: 3780  R: 92  S: 58  HW: 1172  WU: 1785.22  U: 3.82  LW: 2674031
S3/1 Mining BTC at P2Pool: 469.32 Gh/s  (445.72 Gh/s avg)   A: 342  R: 5  S: 0  HW: 40  WU: 6239.09  U: 9.45  LW: 293717
S3/2 Mining BTC at P2Pool: 449.57 Gh/s  (441.23 Gh/s avg)   A: 7839  R: 140  S: 9  HW: 1050  WU: 6176

[philipma1957]

Undervolted two S1's last night once I got S3 online. I guess I did ok. I'm running these two on HP server power supply with breakout board. I have not checked power usage at wall yet.

Clocked @ 250 does that look about right?

S1/1 Mining BTC at P2Pool: 101.56 Gh/s  (121.79 Gh/s avg)   A: 3925  R: 85  S: 34  HW: 5980  WU: 1701.41  U: 3.58  LW: 2889681
S1/2 Mining BTC at P2Pool: 137.68 Gh/s  (127.79 Gh/s avg)   A: 3780  R: 92  S: 58  HW: 1172  WU: 1785.22  U: 3.82  LW: 2674031
S3/1 Mining BTC at P2Pool: 469.32 Gh/s  (445.72 Gh/s avg)   A: 342  R: 5  S: 0  HW: 40  WU: 6239.09  U: 9.45  LW: 293717
S3/2 Mining BTC at P2Pool: 449.57 Gh/s  (441.23 Gh/s avg)   A: 7839  R: 140  S: 9  HW: 1050  WU: 6176

the first s1 looks weak  you are missing hash or info   the hw is 5980  which is okay only if you have a  diffA over 500000

your stats page should list hw diffA and diffR

  error rate = hw/diffA+ diffR      IIRC  so   I can't see your error rate.  my guess is that s-1 is not hitting on all asics  and has an x or 2 in the chain.  the other one is good.

Edit: the error rate is good.   I missed the small print of the time. a hw of 5980 in 18 hours is decent.

  I still think one asic is an x the gh should be a little higher then 121  it should be like your other one 128

try changing to freq 275  to see what happens.  I think the 127 will jump to 140.  not too sure about the one at 121.

you may be able to do one at

 freq 250>> gh of 121-127

  and one at  freq 275>>  gh of 140.

the weaker one may be what you can do at best.  the higher one can stand a bump to freq 275

[squashpile]

You are correct. It is weak and always has been after cold boot especially after undervolt.
Once I power off this one S1 I have to reboot it many times to get good ASIC status. Currently chain two is dorked as usual. Maybe a freq bump will kick it like you mentioned. Usually I can reboot it about 20 times and all is good until next power off. Power off and back on just starts the whole process over on this ant so I have reboot. I got tired of rebooting and left it..lol
xxxxxxoo oooooooo oooooooo oooooooo

Hey I appreciate the info/advice bud.

EDIT: Bumped that good one up to 275. You right man.
140.78 average now

Now back to rebooting the other one..lol

Undervolted two S1's last night once I got S3 online. I guess I did ok. I'm running these two on HP server power supply with breakout board. I have not checked power usage at wall yet.

Clocked @ 250 does that look about right?

S1/1 Mining BTC at P2Pool: 101.56 Gh/s  (121.79 Gh/s avg)   A: 3925  R: 85  S: 34  HW: 5980  WU: 1701.41  U: 3.58  LW: 2889681
S1/2 Mining BTC at P2Pool: 137.68 Gh/s  (127.79 Gh/s avg)   A: 3780  R: 92  S: 58  HW: 1172  WU: 1785.22  U: 3.82  LW: 2674031
S3/1 Mining BTC at P2Pool: 469.32 Gh/s  (445.72 Gh/s avg)   A: 342  R: 5  S: 0  HW: 40  WU: 6239.09  U: 9.45  LW: 293717
S3/2 Mining BTC at P2Pool: 449.57 Gh/s  (441.23 Gh/s avg)   A: 7839  R: 140  S: 9  HW: 1050  WU: 6176

the first s1 looks weak  you are missing hash or info   the hw is 5980  which is okay only if you have a  diffA over 500000

your stats page should list hw diffA and diffR

  error rate = hw/diffA+ diffR      IIRC  so   I can't see your error rate.  my guess is that s-1 is not hitting on all asics  and has an x or 2 in the chain.  the other one is good.

Edit: the error rate is good.   I missed the small print of the time. a hw of 5980 in 18 hours is decent.

  I still think one asic is an x the gh should be a little higher then 121  it should be like your other one 128

try changing to freq 275  to see what happens.  I think the 127 will jump to 140.  not too sure about the one at 121.

you may be able to do one at

 freq 250>> gh of 121-127

  and one at  freq 275>>  gh of 140.

the weaker one may be what you can do at best.  the higher one can stand a bump to freq 275

[bitcoin_miner]

how long time pencil mod will work, if will permanent?

[grn]

how long time pencil mod will work, if will permanent?

until you erase it

[leptar]

how long time pencil mod will work, if will permanent?

keep an eye on errors... the more errors you get the lower the resistance... in my case.. had 2 ants go from .26 to 1.23 in 3 days... the others are holding steady from .09 -.54

7 ants here 3 weeks modded had to touch up 2 ants as the errors increased and the ohms went from 2.88 to 2.45...


I checked the others.. i'm still in the 3.0 range +-.1

[JohanM]

Question about the PCIe connectors and normal 12V-0V connectors.
I am using Corsair CX500M (38A on 12V rail) to power 1 ANt.
When I voltmod them it will be possible to power 2 ants on 1 PSU.

Question.
Can I connect just 1 cable from PSU to ANT, and then connect the 2 normal 12V-0V between the 2 boards ?
Will that work ?
I should think so but never tried...

[lucazane]

Question about the PCIe connectors and normal 12V-0V connectors.
I am using Corsair CX500M (38A on 12V rail) to power 1 ANt.
When I voltmod them it will be possible to power 2 ants on 1 PSU.

Question.
Can I connect just 1 cable from PSU to ANT, and then connect the 2 normal 12V-0V between the 2 boards ?
Will that work ?
I should think so but never tried...

depend on the size of the cable.
if your ant do not need more than 200watts it should be ok.
check if the cable don't get too hot.

[JohanM]

Some experience with undervolting:
- soldering 10k pots is VERY hard. the first 3 took ages since it is both very small and the solder does not take easily. On top of that when you heat too long the smd resistors fry and go up in resistance. luckily you can compensate with the pots.
- the clue is to solder quickly (and use fresh solder within a few secs)
- I used 10k pot in parallel so to lower the resistance to 2.9kohm (from 4.47kohm). I need to set my pots to 8.3kohm for this

So this was stock:
395W at wall (from a Corsair 80+ gold psu), 180Gh/s, 350Mhz

I tried 300Mhz but gave too much HW's.

So then I tried 275Mhz:
192W at wall, 140Gh/s

So a reduction of 23% in hashing power gives a reduction of 52% in power consumption.
Power consumption is now 1.37 W/Gh (down from 2.2 W/Gh).
Fan powers down to 1440rpm (down from 2160rpm)

I'll run this for a day to see how it goes.
Then try 1 antminer with pencil method and see how that holds up.

Anyone did the potentiometer modding with also removing the SMD completely? How is that done ?

[basil3legs]

Anyone did the potentiometer modding with also removing the SMD completely? How is that done ?

The first one I did (OK, so I have only actually done one of the two boards on that one) I used a very hot soldering iron to remove the SMD which was not ideal (pressed on the top mainly as it didn't matter if it got fried).  Also the bit was way too big to solder on the wires and it took ages - hence only the one board done.  I haven't done the other board yet as the pencil is working really well on that one.

After purchasing a much smaller pointy bit I also discovered an old soldering iron I still had and using two to desolder with, one on each end, was really easy!  I was also using a hands free magnifying glass thing which helped a lot.  With the smaller bit, it was also a lot easier to solder the wires on, again with the magnifying glass really helping.

I am currently using c. 0.85V through the chips but am tempted to try lowering it even more and probably will do soon, probably when I get around to soldering the second board on the first one I modded, although I have one other that is still relying just on pencil (and 2 running at full speed still).

One of the two completely modded S1s is running at 275MHz for around 141GH/s at c. 168 Watts at the wall (0.20% errors), the other is running at 262.5MHz for around 135GH/s at c. 160 Watts at the wall (0.65% errors).

[Lucky7Gaming]

I was looking at the antminer S3 page, at it says it is 355 watts with a 225 mhz frequency.....Is it possible that that S3 just has a lot of undervolted S1 chips?   Anyone with a S3 can comment on this?  Or is this a nm change?

[alesx.onfire]

I was looking at the antminer S3 page, at it says it is 355 watts with a 225 mhz frequency.....Is it possible that that S3 just has a lot of undervolted S1 chips?   Anyone with a S3 can comment on this?  Or is this a nm change?

28nm

[crazyates]

I was looking at the antminer S3 page, at it says it is 355 watts with a 225 mhz frequency.....Is it possible that that S3 just has a lot of undervolted S1 chips?   Anyone with a S3 can comment on this?  Or is this a nm change?

That's basically what the S2 was. The S3 actually has fewer chips than a S1, it is in fact a die shrink, and the chips are much faster (per chip).

[Lucky7Gaming]

I was looking at the antminer S3 page, at it says it is 355 watts with a 225 mhz frequency.....Is it possible that that S3 just has a lot of undervolted S1 chips?   Anyone with a S3 can comment on this?  Or is this a nm change?

That's basically what the S2 was. The S3 actually has fewer chips than a S1, it is in fact a die shrink, and the chips are much faster (per chip).

Okay cool.   The s2 was really short lived though. Anyone e have one?

[philipma1957]

good  so you jumped up to 275 and it runs decent.

same idea on the weak one drop it from 250 to 225

see what happens as endless booting should be avoided if you can.

I have found the range on these s-1 and s-3 to be big.  

You are correct. It is weak and always has been after cold boot especially after undervolt.
Once I power off this one S1 I have to reboot it many times to get good ASIC status. Currently chain two is dorked as usual. Maybe a freq bump will kick it like you mentioned. Usually I can reboot it about 20 times and all is good until next power off. Power off and back on just starts the whole process over on this ant so I have reboot. I got tired of rebooting and left it..lol
xxxxxxoo oooooooo oooooooo oooooooo

Hey I appreciate the info/advice bud.

EDIT: Bumped that good one up to 275. You right man.
140.78 average now

Now back to rebooting the other one..lol

Undervolted two S1's last night once I got S3 online. I guess I did ok. I'm running these two on HP server power supply with breakout board. I have not checked power usage at wall yet.

Clocked @ 250 does that look about right?

S1/1 Mining BTC at P2Pool: 101.56 Gh/s  (121.79 Gh/s avg)   A: 3925  R: 85  S: 34  HW: 5980  WU: 1701.41  U: 3.58  LW: 2889681
S1/2 Mining BTC at P2Pool: 137.68 Gh/s  (127.79 Gh/s avg)   A: 3780  R: 92  S: 58  HW: 1172  WU: 1785.22  U: 3.82  LW: 2674031
S3/1 Mining BTC at P2Pool: 469.32 Gh/s  (445.72 Gh/s avg)   A: 342  R: 5  S: 0  HW: 40  WU: 6239.09  U: 9.45  LW: 293717
S3/2 Mining BTC at P2Pool: 449.57 Gh/s  (441.23 Gh/s avg)   A: 7839  R: 140  S: 9  HW: 1050  WU: 6176

the first s1 looks weak  you are missing hash or info   the hw is 5980  which is okay only if you have a  diffA over 500000

your stats page should list hw diffA and diffR

  error rate = hw/diffA+ diffR      IIRC  so   I can't see your error rate.  my guess is that s-1 is not hitting on all asics  and has an x or 2 in the chain.  the other one is good.

Edit: the error rate is good.   I missed the small print of the time. a hw of 5980 in 18 hours is decent.

  I still think one asic is an x the gh should be a little higher then 121  it should be like your other one 128

try changing to freq 275  to see what happens.  I think the 127 will jump to 140.  not too sure about the one at 121.

you may be able to do one at

 freq 250>> gh of 121-127

  and one at  freq 275>>  gh of 140.

the weaker one may be what you can do at best.  the higher one can stand a bump to freq 275

[JohanM]

I finished undervolting my first 3 S1's with 10k pots.
2 are running fine, 1 is getting 1 to 8 x's after 24h.
Does anyone know how to translate the position of the X-marks in the status screen to the actual location on the S1 ??
That'll make it a lot easier to re-check the ant.
I did order a finer point (1.6mm) for my weller to do the next 4 S1 ...

[Bitcoin++]

I'd like to minimize W/GH.

So far I've only tried OP's recommendation. My miners ended up at 185W @ 140 GH = 1.3 W/GH.

Has anyone underclocked even more, and what W/GH do you reach?

EDIT : See OP. He uses 0.85V but shows that 0,75V should be even more efficient. For 0.85V the resistance shall be between ~2.8kOhms and ~3kOhms. We should find out what Ohm that suits the 0.75V setting

[aurel57]

I'd like to minimize W/GH.

So far I've only tried OP's recommendation. My miners ended up at 185W @ 140 GH = 1.3 W/GH.

Has anyone underclocked even more, and what W/GH do you reach?

EDIT : See OP. He uses 0.85V but shows that 0,75V should be even more efficient. For 0.85V the resistance shall be between ~2.8kOhms and ~3kOhms. We should find out what Ohm that suits the 0.75V setting

I have 9 penciled mod down and was thinking of trying to go lower in a couple of weeks. Let me know how it works for you.