BITMAIN Antminer S3 support and OverClocking thread

[no141]

The most telling and important part of the miner is the cgminer version, mine only have v3.12.0-1. Still need to look if I can gain anything by upgrading.

As for voltage, your are stuck. I don't think anyone got these to work at anything higher than 250 and stable.

[Real-Duke]

Wich cgminer version contains the last update from the 9th January ?

[soy]

The most telling and important part of the miner is the cgminer version, mine only have v3.12.0-1. Still need to look if I can gain anything by upgrading.

As for voltage, your are stuck. I don't think anyone got these to work at anything higher than 250 and stable.

No, I'm thinking that I'll absolutely need to lower the wattage/GH/s as the summer heats up and these get more inefficient.  I've seen the thread where resistors are soldered in parallel within the voltage regulator but not prepared to do that.

[no141]

Wich cgminer version contains the last update from the 9th January ?

Current bios uses cgminer v4.6.1. I did not see any direct changes for S3 in the change log from the orignal ver included to the updated one. So, my best guess is it improves communication with some pools.

No, I'm thinking that I'll absolutely need to lower the wattage/GH/s as the summer heats up and these get more inefficient.  I've seen the thread where resistors are soldered in parallel within the voltage regulator but not prepared to do that.

I'll just run a vent tube to the window, to try to help out at hours it's less than 85F outside. It'll be fine running in a closed off room with the rooms vents closed, but that room will be toasty.

[Mikestang]

The file with frequency values for the advanced page with miner settings is

Code:

/usr/lib/lua/luci/model/cbi/cgminer/cgminer.lua

Look at the end of the file for the frequency list.

Here are highlighted in red the lines with new frequency not included in the stock firmware.

pb:value("13:300:0b82", translate("300M"))
pb:value("14:293.75:1706", translate("293.75M"))
pb:value("14:287.5:0b02", translate("287.5M"))
pb:value("14:281.25:1606", translate("281.25M"))
pb:value("15:275:0a82", translate("275M"))
pb:value("15:268.75:1506", translate("268.75M"))
pb:value("15:262.5:0a02", translate("262.5M"))
pb:value("16:256.25:1406", translate("256.25M"))
pb:value("16:250:0982", translate("250M"))
pb:value("16:243.75:1306", translate("243.75M"))
pb:value("17:237.5:1286", translate("237.5M"))
pb:value("17:231.25:1206", translate("231.25M"))
pb:value("18:225:0882", translate("225M"))
pb:value("18:218.75:1106", translate("218.75M (default)"))
pb:value("18:212.5:1086", translate("212.5M"))
pb:value("19:206.25:1006", translate("206.25M"))
pb:value("20:200:0782", translate("200M"))
pb:value("20:196:1f07", translate("196M"))
pb:value("20:193:0f03", translate("193M"))
pb:value("23:175:0d83", translate("175M"))
pb:value("27:150:0b83", translate("150M"))
pb:value("33:125:0983", translate("125M"))
pb:value("40:100:0783", translate("100M"))

Many thanks, just made these changes to my S3+.  Very useful information, kudos to you for providing it!

[valkir]

What is the lowest freq we can go on a S3 ?? For the summer! 

[philipma1957]

What is the lowest freq we can go on a S3 ?? For the summer!  

150 easy.

my batch 1 set of 2 has run 24/7/365 since july 25 of 2014

freq 212  as these did not like to clock high. i can not complain since they have run for just under 9 months so far.

[Mikestang]

After making the changes to my S3+ the new frequencies show up in the gui. I selected 262.5 and input voltage as 0750. When I restarted the box it came back still hashing at 450gh, but the error rate jumped from basically 0 to 0.3%.  I set the frequency then to 250 and left voltage as 0750, restarted, and it came back hashing at 500gh with 0 error again. Why wouldn't the 262.5 frequency work?

The file with frequency values for the advanced page with miner settings is

Code:

/usr/lib/lua/luci/model/cbi/cgminer/cgminer.lua

Look at the end of the file for the frequency list.

Here are highlighted in red the lines with new frequency not included in the stock firmware.

pb:value("13:300:0b82", translate("300M"))
pb:value("14:293.75:1706", translate("293.75M"))
pb:value("14:287.5:0b02", translate("287.5M"))
pb:value("14:281.25:1606", translate("281.25M"))
pb:value("15:275:0a82", translate("275M"))
pb:value("15:268.75:1506", translate("268.75M"))
pb:value("15:262.5:0a02", translate("262.5M"))
pb:value("16:256.25:1406", translate("256.25M"))
pb:value("16:250:0982", translate("250M"))
pb:value("16:243.75:1306", translate("243.75M"))
pb:value("17:237.5:1286", translate("237.5M"))
pb:value("17:231.25:1206", translate("231.25M"))
pb:value("18:225:0882", translate("225M"))
pb:value("18:218.75:1106", translate("218.75M (default)"))
pb:value("18:212.5:1086", translate("212.5M"))
pb:value("19:206.25:1006", translate("206.25M"))
pb:value("20:200:0782", translate("200M"))
pb:value("20:196:1f07", translate("196M"))
pb:value("20:193:0f03", translate("193M"))
pb:value("23:175:0d83", translate("175M"))
pb:value("27:150:0b83", translate("150M"))
pb:value("33:125:0983", translate("125M"))
pb:value("40:100:0783", translate("100M"))

Many thanks, just made these changes to my S3+.  Very useful information, kudos to you for providing it!

[klondike_bar]

After making the changes to my S3+ the new frequencies show up in the gui. I selected 262.5 and input voltage as 0750. When I restarted the box it came back still hashing at 450gh, but the error rate jumped from basically 0 to 0.3%.  I set the frequency then to 250 and left voltage as 0750, restarted, and it came back hashing at 500gh with 0 error again. Why wouldn't the 262.5 frequency work?

0.3% error isnt much, but poor performance doesnt always show as errors - you could need slightly higher voltage

look at the discard rate - if its more than 100% of the accepted sharees, you might have a problem and need to tune lower

[Mikestang]

After making the changes to my S3+ the new frequencies show up in the gui. I selected 262.5 and input voltage as 0750. When I restarted the box it came back still hashing at 450gh, but the error rate jumped from basically 0 to 0.3%.  I set the frequency then to 250 and left voltage as 0750, restarted, and it came back hashing at 500gh with 0 error again. Why wouldn't the 262.5 frequency work?

0.3% error isnt much, but poor performance doesnt always show as errors - you could need slightly higher voltage

look at the discard rate - if its more than 100% of the accepted sharees, you might have a problem and need to tune lower

My main point was that the increased frequency and voltage did not increase the hash rate, it should have been hashing ~530gh, but it was hashing like at stock 225 freq.

[Moria843]

I can't believe there's still so much interest in overclocking. Aren't you overclockers looking at the effeciency with respect to energy cost? Unless you're getting free electricity it's not worth it, especially with warmer weather coming to us in the northern hemisphere.

I'm wanting to underclock before summer and have seen many discussions related on which firmware you need and how to do. But, have seen a lot of conflicting info.

Is there a good S3 underclocking thread anywhere or a "solved or proven/best" technique?

[Mikestang]

Aren't you overclockers looking at the effeciency with respect to energy cost?

Nope, I mine for hobby, I want as much hash as I can get while keeping hardware errors under 1%.  Heat and electricity cost be damned! 

[sloopy]

After making the changes to my S3+ the new frequencies show up in the gui. I selected 262.5 and input voltage as 0750. When I restarted the box it came back still hashing at 450gh, but the error rate jumped from basically 0 to 0.3%.  I set the frequency then to 250 and left voltage as 0750, restarted, and it came back hashing at 500gh with 0 error again. Why wouldn't the 262.5 frequency work?

0.3% error isnt much, but poor performance doesnt always show as errors - you could need slightly higher voltage

look at the discard rate - if its more than 100% of the accepted sharees, you might have a problem and need to tune lower

All of my miners have a much higher than 100% discard rate versus accepted.
The SP20s, S3s, S4s

What do you have on yours?

[CapnBDL]

Howdy,

Dang it...my eyes hurt!!  Somewhere I read about changing the 'queue' size to help with so many discards. Can anybody help? Ideas about a good resize number and the file that needs to be edited?

Thank you ever so much.

[no141]

I noticed a few interesting lines in the logs. I don't remember anything happening when it occurred, just wondering if it's due to the overclock or anything to worry about. Here' the lines:

[383343.070000] bitmain 1-1.1:1.0: USB Bitmain asic device now attached to USB Bitmain asic-0
[383343.080000] PIC microchip usb OK
[383343.910000] bitmain 1-1.1:1.0: USB Bitmain asic #0 now disconnected

That might be what it does when one of the cores has an issue? I see chain 2 had oooxoooo oooooooo, so I rebooted it and it's fine again. Looks like I'll need to take it all apart and fix up the thermal paste on the chips.

[no141]

Howdy,

Dang it...my eyes hurt!!  Somewhere I read about changing the 'queue' size to help with so many discards. Can anybody help? Ideas about a good resize number and the file that needs to be edited?

Thank you ever so much.

When something goes wrong, or so I don't have to log into luci, I just run cgminer in ssh. Once you kill the cgminer that's already running, you can just start it up again in the shell. Once you have it started, that's the cgminer the web interface will monitor and that's what the cron job will take care of if something happens. If I didn't leave a computer on, I'd just run the miner to screen session or just a file my router. That way you can see everything that happens and test things too. While you are looking for the stuff to edit on the miner, test things out simply.

Here's my line for a 250 overclock:

cgminer --bitmain-options 115200:32:8:16:250:0982 -o pool1 -O user:pass -o pool2 -O user:pass -o pool3 -O user:pass --api-listen --api-network --bitmain-checkn2diff --bitmain-hwerror --queue 4096

See the queue at the end? Edit that and test things. As a bonus, if you want to list more pools than 3, just keep adding them in the command-line. You'll want to copy and paste your line.

If and when you are ready, either add a new config item in the /etc/config/cgminer file for queue size, then change the static entry in /etc/init.d/cgminer (along with a config_get) to the variable one. Or just edit the init.d file with the queue size you want. Yep, I leave the easy way last.

[CapnBDL]

Howdy,

Dang it...my eyes hurt!!  Somewhere I read about changing the 'queue' size to help with so many discards. Can anybody help? Ideas about a good resize number and the file that needs to be edited?

Thank you ever so much.

When something goes wrong, or so I don't have to log into luci, I just run cgminer in ssh. Once you kill the cgminer that's already running, you can just start it up again in the shell. Once you have it started, that's the cgminer the web interface will monitor and that's what the cron job will take care of if something happens. If I didn't leave a computer on, I'd just run the miner to screen session or just a file my router. That way you can see everything that happens and test things too. While you are looking for the stuff to edit on the miner, test things out simply.

Here's my line for a 250 overclock:

cgminer --bitmain-options 115200:32:8:16:250:0982 -o pool1 -O user:pass -o pool2 -O user:pass -o pool3 -O user:pass --api-listen --api-network --bitmain-checkn2diff --bitmain-hwerror --queue 4096

See the queue at the end? Edit that and test things. As a bonus, if you want to list more pools than 3, just keep adding them in the command-line. You'll want to copy and paste your line.

If and when you are ready, either add a new config item in the /etc/config/cgminer file for queue size, then change the static entry in /etc/init.d/cgminer (along with a config_get) to the variable one. Or just edit the init.d file with the queue size you want. Yep, I leave the easy way last.

Ya know...you would. Such a d***. lol I'm kidding,.. of course. Jeez...I do love a person with humor. Ya put me thru all the blah, blah, which I have no clue about, then... Did get a chuckle tho. Ok..I get that I want to edit the init.d, but, I'm not a pro and can't find the d*** thing. Plus, not great with puTTy or code. Sure could use a hand and not clues. Comon. ..I wanna play...not work. How 'bout it? Help an ol' man out. BTW...thanks. (first things first). Later...how do ya run cgminer in ssh and is that upgradeable? Ohhhh...run with no computer?..load into the router? FASCINATING! Hmmmm... Could be interesting,..wanna know about the init first.
Thank you

[no141]

Howdy,

Dang it...my eyes hurt!!  Somewhere I read about changing the 'queue' size to help with so many discards. Can anybody help? Ideas about a good resize number and the file that needs to be edited?

Thank you ever so much.

When something goes wrong, or so I don't have to log into luci, I just run cgminer in ssh. Once you kill the cgminer that's already running, you can just start it up again in the shell. Once you have it started, that's the cgminer the web interface will monitor and that's what the cron job will take care of if something happens. If I didn't leave a computer on, I'd just run the miner to screen session or just a file my router. That way you can see everything that happens and test things too. While you are looking for the stuff to edit on the miner, test things out simply.

Here's my line for a 250 overclock:

cgminer --bitmain-options 115200:32:8:16:250:0982 -o pool1 -O user:pass -o pool2 -O user:pass -o pool3 -O user:pass --api-listen --api-network --bitmain-checkn2diff --bitmain-hwerror --queue 4096

See the queue at the end? Edit that and test things. As a bonus, if you want to list more pools than 3, just keep adding them in the command-line. You'll want to copy and paste your line.

If and when you are ready, either add a new config item in the /etc/config/cgminer file for queue size, then change the static entry in /etc/init.d/cgminer (along with a config_get) to the variable one. Or just edit the init.d file with the queue size you want. Yep, I leave the easy way last.

Ya know...you would. Such a d***. lol I'm kidding,.. of course. Jeez...I do love a person with humor. Ya put me thru all the blah, blah, which I have no clue about, then... Did get a chuckle tho. Ok..I get that I want to edit the init.d, but, I'm not a pro and can't find the d*** thing. Plus, not great with puTTy or code. Sure could use a hand and not clues. Comon. ..I wanna play...not work. How 'bout it? Help an ol' man out. BTW...thanks. (first things first). Later...how do ya run cgminer in ssh and is that upgradeable? Ohhhh...run with no computer?..load into the router? FASCINATING! Hmmmm... Could be interesting,..wanna know about the init first.
Thank you

Well, I know enough linux to make what I said child's play, so here's a few things to know:
cd /etc - will get you in the etc dir for example
ls - gets a list out of the files and dirs (I like 'ls -la' for more detail)
ps w - lists all processes
kill [pid of cgminer] - kills the running program as seen in ps
   Here's the relevant entry that ps showed me:  1169 root     49812 S <  cgminer --bitmain-options 115200:32:8:16:250:0982 -o stratum...
   It's pid is that first number or 1169 in this case.
cat [file] - types out the file you specify
vi - see the first post in this thread. 'i' starts edit mode, Esc will stop edit mode, 'ZZ' will save and exit when not in edit mode.

The miner on the latest update uses BusyBox v1.19.4. You could use that as a starting point to learn more what the linux on the machine can do.

If you only play with the cgminer files, you can only mess up that program running. So I'd suggest putting a # in front of lines that you will be editing, then copy the line in a new line with your changes. Or remember what you did--seems to make the most sense as you are worried about the queue.


I gave you the line to run cgminer in ssh already. You would just need to replace the pools and user:pass with real info.
If you add the patched cgminer, like from nicehash, you could run either version, as long as you renamed the original first. Other than that patch, the newer versions offer no additional speed benefits to a S3.

When summer rates kick in or it gets too hot, I'll probably have the main computer sleep or hibernate. In that case, the router will always be on, so I ssh to that from my phone, and then I use a script that just basically has the command line in it. The router is flashed with a version of Tomato, so that makes it a useful linux box. I use flash drives or a hard disk with it depending on how much storage it needs for what I need it to do.

[no141]

i explained all this on this sight but no one would listen.

the firmware is screwed up and if you don't uncheck the save settings box above your flashing file  it will crash during flash and will be unbootable.
i explained this to many people the other day and no one believed me thy thought i was lying. but now you are down.
i have 26 s3's and my s3' were crashing many times a day thy were working good and then crash down to 40 gh/s i have been working on this problem for a month and finely fixed them.
if you had flashed bitmains last firmware 1226 thats the problem and reset wont work and if you flash back to earlier program it crashes the miner and is un bootable.
i am willing to enplane step by step how to recover the unbootable miner and flash crashing miners back to good firmware with success. bitmain is making tones of btc telling people the control boards are bad just like thy told me and i fixed it. it took me three hrs to set up and figure it out and now that im setup it only takes 15 minuts to fix a crashed s3's.
if you want the help i can give it and in return it would be nice to get a btc donation for saving you tons of time and money
but it can still be fixed

Hey read your post was trying to downgrade my firmware and I stopped in the middle flashing and now the red light blinks and cant reset. What is the solution ill send some btc your way if it works.

I think it's hard to brick these. First did you try holding down the reset button down when the unit was on for one minute and then having it pressed down for about a minute from when it was off powering it up?

The method generally available to devices essentially in panic mode is TFTP, and it may have a very limited shell to recover with. I'm surprised proper step-by-step info for Antminers with bad flashes can't be easily found here, instead of holding a pricetag on it. You will need to find out if it manages to connect to the network(wired only) and then do a port scan to see what low numbered ports are open if you want to see it it supports either.

[luthermarcus]

Dont know what happen but i was downgrading to older firmware. I did exactly like it said on forum i mad sure that I unchecked keep settings wanted a clean flash but it booted up with red light blinking cant connect to miner isp log on. I made sure i put in the proper isp. Reset wont work there is no reason to believe that i fried it becuase i was reflashing and it was running fine, I spent 3 hours trying every possible way to reset hardware nothing its not the first time i reset it it just wont work. I was thinking is i could attach the unit to run of the S4 controller board connected to its own power supply and linking the chip boards.