BITMAIN Antminer S3 support and OverClocking thread

[vulgartrendkill]

Is there a go-to power supply any of you have been going to power 2-4 of these units?
Most seem limited on PCI-e cables, I would love to know.

I'm using an OCZ 1000W ZX. Can cope with a slight overclock looking at the numbers, but as yet unable to verify.  Runs perfect on stock. Has 6 PCIE cables.

[Beastlymac]

Strangely on both of mine. their conf is
option 'freq_value'    '0982'  #225M  in comment instead of #250
option 'chip_freq'     '250'
option 'timeout'       '16'

The 225 settings are (currently using this setting @ 453 gh/s avg from 440gh/s at #218.75M)

option 'freq_value'    '0882'  #225M
option 'chip_freq'     '225'
option 'timeout'       '16'

I was looking for the #250 option which wasnt there and then i spotted the chip freq value of 250 in the #225 settings

Same happened with me. Seems to just be a type o on a very early version of the software.

[dlasher]

Quick power #'s, measured with my best unit.. and the only one that will hit 500GH.

Corsair TX850 80plus bronze PSU - single S3:
218 mhz -- 450 GH --  346 watt at the wall
225 mhz -- 480 GH --  370 watt at the wall
237 mhz -- 500 GH --  394 watt at the wall

My first order was 10, and here's an average over the last few hours:

miner01   501.87 GH/s   3404032 (99.91%)   3072 / 0 / 0   0:00:03   
miner02   456.70 GH/s   3092736 (99.93%)   2304 / 0 / 0   0:00:03   
miner03   422.43 GH/s   3056128 (99.88%)   3584 / 0 / 0   0:00:03   
miner04   442.37 GH/s   3094016 (99.62%)   11648 / 0 / 0   0:00:03   
miner05   423.99 GH/s   3081984 (99.87%)   4096 / 0 / 0   0:00:03   
miner06   429.29 GH/s   3043328 (99.68%)   9728 / 0 / 0   0:00:03   
miner07   430.22 GH/s   2934272 (99.64%)   10496 / 0 / 0   0:00:04   
miner08   451.09 GH/s   3077632 (99.88%)   3584 / 0 / 0   0:00:04   
miner09   441.12 GH/s   2973696 (99.51%)   14592 / 0 / 0   0:00:04   
miner10   454.52 GH/s   3096832 (99.69%)   9728 / 0 / 0   0:00:04

Overall, kinda disappointed with the lack of overclocking. (Which was warned about due to a bad batch of DC-DC..)  While I do have one fast unit, One of the units won't even run at the STOCK clocks without X'ing out at least 1-2 of the ASIC's, and most of them can't survive even bumping up to 225. At this stage 2 or 4 power connectors doesn't seem to matter at all. The 500GH unit is running on (2), and the 420-units won't go faster on (4). I'm running 3 S3's per 1300W power supply, so it's not a power issue.

Feeling a little deju'vu to KNC here.. First batch got bad dies, 400GH, second batch got fixes, 550GH.

[J4bberwock]

For the ones who'd like to try other frequencies, they are in the data sheet available on bitmain site.
Or maybe the old table from the antminer over clocking thread would work after swaping the last digit from "1" to "2".
It seems to be the only difference between the settings.
https://bitcointalk.org/index.php?topic=348327.msg7544198#msg7544198

Direct link to data sheet

https://bitmaintech.com/files/download/BM1382_Datasheet_v3.0.pdf

I guess we can calculate the acceptable frequencies the same way as in the old thread, but I'm on my phone, so not that easy.

[dlasher]

For the ones who'd like to try other frequencies, they are in the data sheet available on bitmain site.
Or maybe the old table from the antminer over clocking thread would work after swaping the last digit from "1" to "2".
It seems to be the only difference between the settings.

Link?

[J4bberwock]

For the ones who'd like to try other frequencies, they are in the data sheet available on bitmain site.
Or maybe the old table from the antminer over clocking thread would work after swaping the last digit from "1" to "2".
It seems to be the only difference between the settings.

Link?

You are too fast, I was editing

[dlasher]

Anyone have packages available yet with a newer cgminer? S3's come with 3.1.2.. old and crusty...

(Or instructions on how to get a build/dev/chain working, and I'll compile my own on something with horsepower)

[finlof]

i did my own calculations and here are the specs for 243.75M:

        option 'freq_value'    '1306'  #243.75M
        option 'chip_freq'     '243.75'
        option 'timeout'       '17'

[dlasher]

generated a table to work with, based on the S1/S2 tables:

...package 'cgminer'

config 'asic-freq' 'default'

        #option 'freq_value'    '1486'  #262.5M
        #option 'chip_freq'     '262.5'
        #option 'timeout'       '15'

        #option 'freq_value'    '1386'  #250M
        #option 'chip_freq'     '250'
        #option 'timeout'       '16'

        #option 'freq_value'    '1306'  #243.75M
        #option 'chip_freq'     '243.75'
        #option 'timeout'       '17'

        #option 'freq_value'    '1286'  #237.5M
        #option 'chip_freq'     '237.5'
        #option 'timeout'       '17'

        #option 'freq_value'    '1206'  #231.25M   
        #option 'chip_freq'     '231.5'             
        #option 'timeout'       '18'                                                         
                                                                     
        #option 'freq_value'    '1185'  #225M   
        #option 'chip_freq'     '225'           
        #option 'timeout'       '18'           
                                               
        option 'freq_value'    '1106'  #218.75M
        option 'chip_freq'     '218.75'       
        option 'timeout'       '18'           
                                               
        #option 'freq_value'    '1086'  #212.5M
        #option 'chip_freq'     '212.5'       
        #option 'timeout'       '18'         
                                             
        #option 'freq_value'    '0782'  #200M
        #option 'chip_freq'     '200'       
        #option 'timeout'       '20'

[TwoPints]

Quick power #'s, measured with my best unit.. and the only one that will hit 500GH.

Corsair TX850 80plus bronze PSU - single S3:
218 mhz -- 450 GH --  346 watt at the wall
225 mhz -- 480 GH --  370 watt at the wall
237 mhz -- 500 GH --  394 watt at the wall

I was a bit curious myself and tested one of my batch one units, but I'm getting different power numbers at the wall. See below:

218 mhz -- 440 GH --  360 watt at the wall
225 mhz -- 387 GH --  348 watt at the wall
237 mhz -- 408 GH --  360 watt at the wall
250 mgh -- 424 GH --  345 watt at the wall

I'm using a Corsair HX650 for each unit and testing using a Rosewill RHSP-13001. I've only tested one S3 thus far but there appears to be a power issue, or it's just not liking anything outside of stock frequency. I haven't flashed the firmware yet but am more concerned that I'm not pushing the same power to the unit at those higher frequencies. Any thoughts?

[jamesc760]

generated a table to work with, based on the S1/S2 tables:

...package 'cgminer'

config 'asic-freq' 'default'

        #option 'freq_value'    '1486'  #262.5M
        #option 'chip_freq'     '262.5'
        #option 'timeout'       '15'

        #option 'freq_value'    '1386'  #250M
        #option 'chip_freq'     '250'
        #option 'timeout'       '16'

        #option 'freq_value'    '1306'  #243.75M
        #option 'chip_freq'     '243.75'
        #option 'timeout'       '17'

        #option 'freq_value'    '1286'  #237.5M
        #option 'chip_freq'     '237.5'
        #option 'timeout'       '17'

        #option 'freq_value'    '1206'  #231.25M   
        #option 'chip_freq'     '231.5'             
        #option 'timeout'       '18'                                                         
                                                                     
        #option 'freq_value'    '1185'  #225M   
        #option 'chip_freq'     '225'           
        #option 'timeout'       '18'           
                                               
        option 'freq_value'    '1106'  #218.75M
        option 'chip_freq'     '218.75'       
        option 'timeout'       '18'           
                                               
        #option 'freq_value'    '1086'  #212.5M
        #option 'chip_freq'     '212.5'       
        #option 'timeout'       '18'         
                                             
        #option 'freq_value'    '0782'  #200M
        #option 'chip_freq'     '200'       
        #option 'timeout'       '20'

Thank you!

Is there a setting for 220M?

[dlasher]

Quick power #'s, measured with my best unit.. and the only one that will hit 500GH.

Corsair TX850 80plus bronze PSU - single S3:
218 mhz -- 450 GH --  346 watt at the wall
225 mhz -- 480 GH --  370 watt at the wall
237 mhz -- 500 GH --  394 watt at the wall

I was a bit curious myself and tested one of my batch one units, but I'm getting different power numbers at the wall. See below:

218 mhz -- 440 GH --  360 watt at the wall
225 mhz -- 387 GH --  348 watt at the wall
237 mhz -- 408 GH --  360 watt at the wall
250 mgh -- 424 GH --  345 watt at the wall

I'm using a Corsair HX650 for each unit and testing using a Rosewill RHSP-13001. I've only tested one S3 thus far but there appears to be a power issue, or it's just not liking anything outside of stock frequency. I haven't flashed the firmware yet but am more concerned that I'm not pushing the same power to the unit at those higher frequencies. Any thoughts?

So in the apology post that Bitmain posted, ( https://bitcointalk.org/index.php?topic=671189.msg7832797#msg7832797 ) they effectively indicated a bad batch of DC-DC and people weren't going to get the speeds they wanted.. hence the refund/credit.

I suspect they've set all S3's to 218 out of the box, and knowing that most of them won't be able to run over that point. I have (1) of (10) that can run over the stock.. the rest just start X'ing out ASIC's and never go any better.. they're actually faster are 218 than at 225/237/250/etc.

(I know a couple of -really- good board-level guys who could swap-out DC-DC supplies if I had any idea what to replace the faulty ones with)

[J4bberwock]

Quick power #'s, measured with my best unit.. and the only one that will hit 500GH.

Corsair TX850 80plus bronze PSU - single S3:
218 mhz -- 450 GH --  346 watt at the wall
225 mhz -- 480 GH --  370 watt at the wall
237 mhz -- 500 GH --  394 watt at the wall

I was a bit curious myself and tested one of my batch one units, but I'm getting different power numbers at the wall. See below:

218 mhz -- 440 GH --  360 watt at the wall
225 mhz -- 387 GH --  348 watt at the wall
237 mhz -- 408 GH --  360 watt at the wall
250 mgh -- 424 GH --  345 watt at the wall

I'm using a Corsair HX650 for each unit and testing using a Rosewill RHSP-13001. I've only tested one S3 thus far but there appears to be a power issue, or it's just not liking anything outside of stock frequency. I haven't flashed the firmware yet but am more concerned that I'm not pushing the same power to the unit at those higher frequencies. Any thoughts?

So in the apology post that Bitmain posted, ( https://bitcointalk.org/index.php?topic=671189.msg7832797#msg7832797 ) they effectively indicated a bad batch of DC-DC and people weren't going to get the speeds they wanted.. hence the refund/credit.

I suspect they've set all S3's to 218 out of the box, and knowing that most of them won't be able to run over that point. I have (1) of (10) that can run over the stock.. the rest just start X'ing out ASIC's and never go any better.. they're actually faster are 218 than at 225/237/250/etc.

(I know a couple of -really- good board-level guys who could swap-out DC-DC supplies if I had any idea what to replace the faulty ones with)

As far as I know, there are no pin to pin replacement for the 53355 with same or higher amp in the same voltage values.
I already checked for overvolting.
Best bet would be to replace it with another batch of 53355, 3$ each, 50$ for each S3, and not sure it will solve anything.
for low cost fix, I'd try a slight over voltage by replacing the divider resistor. Going up to 0.8v could help.

[klondike_bar]

of 4 units:

a) 231.25MHZ = 452Gh
b) 218.75MHz = 429GH (one chip is shown as -  though, so presumably im losing about 12GH)
c) 250MHz = 502.5GH
d) 218.75MHz = 428GH (one chip is shown as -  though, so presumably im losing about 12GH)

ive tried rebooting the two systems with the dudding chips but to no avail. has anyone had and resolved this issue? (with the S1 a reboot always seemed to solve it)

overall im not terribly impressed with the quality of the components bitmain used. hopefully its a batch-specific problem though and wont occur again.

[daddyfatsax]

of 4 units:

a) 231.25MHZ = 452Gh
b) 218.75MHz = 429GH (one chip is shown as -  though, so presumably im losing about 12GH)
c) 250MHz = 502.5GH
d) 218.75MHz = 428GH (one chip is shown as -  though, so presumably im losing about 12GH)

ive tried rebooting the two systems with the dudding chips but to no avail. has anyone had and resolved this issue? (with the S1 a reboot always seemed to solve it)

overall im not terribly impressed with the quality of the components bitmain used. hopefully its a batch-specific problem though and wont occur again.

I had a unit that would X chips at any clock about 218.75. Had to take it apart and reapply thermal paste. No X chips after that. Been running for 2 days on stock clock with no issues.

[Azmodeus]

of 4 units:

a) 231.25MHZ = 452Gh
b) 218.75MHz = 429GH (one chip is shown as -  though, so presumably im losing about 12GH)
c) 250MHz = 502.5GH
d) 218.75MHz = 428GH (one chip is shown as -  though, so presumably im losing about 12GH)

ive tried rebooting the two systems with the dudding chips but to no avail. has anyone had and resolved this issue? (with the S1 a reboot always seemed to solve it)

overall im not terribly impressed with the quality of the components bitmain used. hopefully its a batch-specific problem though and wont occur again.

I had a unit that would X chips at any clock about 218.75. Had to take it apart and reapply thermal paste. No X chips after that. Been running for 2 days on stock clock with no issues.

When you guys are re-applying thermal paste after cleaning with alchohol are you applying a thin layer with a piece of plastic/razor so the entire chip is covered or do you just put like a 1/2 pea sized drop in the center of each chip and let the heatsink smoosh it around?

[bytechanger]

When you guys are re-applying thermal paste after cleaning with alchohol are you applying a thin layer with a piece of plastic/razor so the entire chip is covered or do you just put like a 1/2 pea sized drop in the center of each chip and let the heatsink smoosh it around?

Technically, the 2nd procedure you mentioned is correct.  Applying it w/ plastic or razor has the potential to create micro air bubbles in the paste and hinder its ability to transfer heat away from the chip.

I would test a small amount on one chip, and see what the coverage is like.

Once you have roughly the ideal amount, simply repeat the process down the line...

A little dab will usually suffice...

[daddyfatsax]

When you guys are re-applying thermal paste after cleaning with alchohol are you applying a thin layer with a piece of plastic/razor so the entire chip is covered or do you just put like a 1/2 pea sized drop in the center of each chip and let the heatsink smoosh it around?

Technically, the 2nd procedure you mentioned is correct.  Applying it w/ plastic or razor has the potential to create micro air bubbles in the paste and hinder its ability to transfer heat away from the chip.

I would test a small amount on one chip, and see what the coverage is like.

Once you have roughly the ideal amount, simply repeat the process down the line...

A little dab will usually suffice...

Yeah a little dab is all you need. This is not like a CPU where you can put a 1/2 of a pea worth and let the heat sink spread it out. Just a little dab on the chip and do not smooth it out. Worked well for me.

[Junkbarman]

After changing the # to where I wanted them, I had to hit esc key and then :wq. I'm a win7 user, and this worked for me.

[notlist3d]

Well batch one here.  One runs at 420-425 overclocking makes no difference.  The other one i am able to get with 238 freq 440-460.  Seems to go down after a few hours.  Didn't try 250 freq as I am happy with the 440+ with temps under 40.  It's very very stable almost 2 days at overclocked speeds on 2nd unit no problems.

Very happy with my unit's wish first one overclocked, but with them going to refund/coupon for speed difference I can't complain at all.