[DIY] - Reward $100 | Antminer S1/S3 Blade on Raspberry Pi - page 40.

pekatete

hero member

Activity: 518

Merit: 500

Quote from: J4bberwock on September 12, 2014, 10:35:57 AM

Quote from: pekatete on September 12, 2014, 09:42:03 AM

I added a third edit - those values were for an S3 here's the one from an S1 --bitmain-options 115200:32:8:35:400:4f81 and they are from the System Log NOT the Kernel Log

For some reasons, touching the baudrate messes the hashrate.

cgminer.exe -o mypool:port -u myworker -p x --bmsc-freq 0F81

gives almost 60Gh on one panel, but still HW errors.
And the chips are all hot.

The cgminer for U2 is quite different, and the parameters aren't the same.

I think you are moving in the right direction and I probably should not ask this, but do you have the fans blowing full pelt seing you have bypassed the temp control?
Yes, the parameters are clearly different, but at least there's one for setting the hex value (as per your line), now what about the freq value? Any ideas what the parameter is?
Come to think of it, I think the time is ripe for us to either:
1. request the maintainer of cgminer incorporate the S1 (and S3) drivers into a windows binary or:
2. for us to do our own compilation to include those drivers.

as clearly the U2 compilation is going to give us headaches galore!

J4bberwock

hero member

Activity: 840

Merit: 1000

Quote from: pekatete on September 12, 2014, 09:42:03 AM

I added a third edit - those values were for an S3 here's the one from an S1 --bitmain-options 115200:32:8:35:400:4f81 and they are from the System Log NOT the Kernel Log
[/quote]
For some reasons, touching the baudrate messes the hashrate.

cgminer.exe -o mypool:port -u myworker -p x --bmsc-freq 0F81

gives almost 60Gh on one panel, but still HW errors.
And the chips are all hot.

The cgminer for U2 is quite different, and the parameters aren't the same.

tarmi

legendary

Activity: 1232

Merit: 1011

Quote from: chadwickx16 on September 12, 2014, 09:56:24 AM

~~Where do you get the UART to USB Converters from?~~ Never mind, I forgot the Google is my friend sometimes. Roll Eyes

Great work J4bberwock!!!!!

What is the male to male jumper called between the cable coming from the USB and into the Ant's cable? Sorry UberN00b here.

you can get the converter and cable for 5 $ - >

http://www.dx.com/p/usb-to-uart-5-pin-cp2102-module-serial-converter-81872

chadwickx16

hero member

Activity: 826

Merit: 1000

~~Where do you get the UART to USB Converters from?~~ Never mind, I forgot the Google is my friend sometimes. Roll Eyes

Great work J4bberwock!!!!!

Will this one will work??

What is the male to male jumper called between the cable coming from the USB and into the Ant's cable? Sorry UberN00b here, never really attempted anything like this before.

smit1237

full member

Activity: 182

Merit: 100

May the source be with you.

omg still no luck. Blade blinks but nothing happen Sad

pekatete

hero member

Activity: 518

Merit: 500

Quote from: J4bberwock on September 12, 2014, 09:40:06 AM

Quote from: pekatete on September 12, 2014, 09:31:20 AM

Quote from: J4bberwock on September 12, 2014, 09:20:45 AM

Here is the test setup.
Getting 45 Gh from a board, but many HW errors

Brilliant!
1. The errors I think will have to do with the driver in cgminer. Look at setting these values in the initialisation string of cgminer, I got this from the Kernel Log of a running miner:

Code:

 --bitmain-options 115200:32:8:15:262.5:0a02 -o stratum+tcp://my.pool.io:3333 -O username:password -o stratum+tcp://my.backup:3333 -O username:password -o stratum+tcp://my.otherbackup.com:9332 -O username:password --api-listen --api-network --bitmain-checkn2diff --bitmain-hwerror --queue 10 --scan-time 1 --expiry 1

EDIT: I think this bit --bitmain-options 115200:32:8:15:262.5:0a02 could be helpful.
EDIT2: That sets the frequency to 262.5 and HEX value to 0a02 - replace accordingly.

2. Do all the chips feel warm to the touch? I have a feeling you are not interfacing with all the chips.

yes, they are all hot/warm, but the default frequency for the U2 is lower than the one on S1.
I'll test your settings and report back

I added a third edit - those values were for an S3 here's the one from an S1 --bitmain-options 115200:32:8:35:400:4f81 and they are from the System Log NOT the Kernel Log

J4bberwock

hero member

Activity: 840

Merit: 1000

Quote from: pekatete on September 12, 2014, 09:31:20 AM

Quote from: J4bberwock on September 12, 2014, 09:20:45 AM

Here is the test setup.
Getting 45 Gh from a board, but many HW errors

Brilliant!
1. The errors I think will have to do with the driver in cgminer. Look at setting these values in the initialisation string of cgminer, I got this from the Kernel Log of a running miner:

Code:

 --bitmain-options 115200:32:8:15:262.5:0a02 -o stratum+tcp://my.pool.io:3333 -O username:password -o stratum+tcp://my.backup:3333 -O username:password -o stratum+tcp://my.otherbackup.com:9332 -O username:password --api-listen --api-network --bitmain-checkn2diff --bitmain-hwerror --queue 10 --scan-time 1 --expiry 1

EDIT: I think this bit --bitmain-options 115200:32:8:15:262.5:0a02 could be helpful.
EDIT2: That sets the frequency to 262.5 and HEX value to 0a02 - replace accordingly.

2. Do all the chips feel warm to the touch? I have a feeling you are not interfacing with all the chips.

yes, they are all hot/warm, but the default frequency for the U2 is lower than the one on S1.
I'll test your settings and report back

pekatete

hero member

Activity: 518

Merit: 500

Quote from: J4bberwock on September 12, 2014, 09:20:45 AM

Here is the test setup.
Getting 45 Gh from a board, but many HW errors

Brilliant!
1. The errors I think will have to do with the driver in cgminer. Look at setting these values in the initialisation string of cgminer, I got this from the Kernel Log of a running miner:

Code:

 --bitmain-options 115200:32:8:15:262.5:0a02 -o stratum+tcp://my.pool.io:3333 -O username:password -o stratum+tcp://my.backup:3333 -O username:password -o stratum+tcp://my.otherbackup.com:9332 -O username:password --api-listen --api-network --bitmain-checkn2diff --bitmain-hwerror --queue 10 --scan-time 1 --expiry 1

EDIT: I think this bit --bitmain-options 115200:32:8:15:262.5:0a02 could be helpful.
EDIT2: That sets the frequency to 262.5 and HEX value to 0a02 - replace accordingly.
EDIT3: HANGON! Those values I got from my S3, heres one from the S1 --bitmain-options 115200:32:8:35:400:4f81

2. Do all the chips feel warm to the touch? I have a feeling you are not interfacing with all the chips.

J4bberwock

hero member

Activity: 840

Merit: 1000

Here is the test setup.
Getting 45 Gh from a board, but many HW errors

J4bberwock

hero member

Activity: 840

Merit: 1000

Quote from: pekatete on September 12, 2014, 08:45:32 AM

Yep, lots of permutations, like I said, the old lady still has some juice left in her!
Did you use one of these?

Yes, a simple one like this.
Costs 3$ or so, but it will be cleaner to have a nice board with integrated usb hub so you can plug a few hashing boards on it.
I'll work on it this weekend.

powersup

sr. member

Activity: 406

Merit: 250

Quote from: bluecityste on September 12, 2014, 08:48:30 AM

So, does this mean (in theory) that this system can be used to control multiple S3 blades from the S1 upgrade kits?

Well done J4bberwock (i think! Unfortunately most of this thread is way aver my head).

Stephen

In theory yes, but it would require some heavy mod work. The problem is the blades are manufactured identically. Each chip on the blade is assigned an address on the UART bus. The problem with multiple blades on the same bus is there will be address conflicts. The MCU on the S1 and S3 controller uses 2 UART ports, one for each blade.

Why is say in theory it can be done is because you can change the address of the chips by bridging pins, so, in theory you could change the chip addresses on the second and third blade and have them all run off the same UART bus.

J4bberwock

hero member

Activity: 840

Merit: 1000

Quote from: smit1237 on September 12, 2014, 08:54:02 AM

Damn i can't get it to work Angry

Looks like i have some bad wiring

pin3 is Rx, pin5 is tx
Pin 1 in the closest to the fan connector.
Doesn't work for me with the newest bfgminer.
Cgminer from december in bitmain's github works ok without any parameter except the pool settings

J4bberwock

hero member

Activity: 840

Merit: 1000

Quote from: bluecityste on September 12, 2014, 08:48:30 AM

So, does this mean (in theory) that this system can be used to control multiple S3 blades from the S1 upgrade kits?

Well done J4bberwock (i think! Unfortunately most of this thread is way aver my head).

Stephen

Yes, but the cgminer part still needs to be fixed correctly, I'm not a coder, but I can have a look, it will be slower than if someone actually knowing what to do is looking at it.

I'm currently having high HW errors because I can't choose the frequency and the timeout, and maybe another parameter.
In the december build, it looks like the values are hardcoded, and the newer version gives me even worst results for now.

smit1237

full member

Activity: 182

Merit: 100

May the source be with you.

Damn i can't get it to work Angry

Looks like i have some bad wiring

bluecityste

newbie

Activity: 47

Merit: 0

So, does this mean (in theory) that this system can be used to control multiple S3 blades from the S1 upgrade kits?

Well done J4bberwock (i think! Unfortunately most of this thread is way aver my head).

Stephen

pekatete

hero member

Activity: 518

Merit: 500

Quote from: J4bberwock on September 12, 2014, 08:34:47 AM

Quote from: pekatete on September 12, 2014, 08:14:15 AM

Quote from: J4bberwock on September 12, 2014, 08:06:14 AM

The control board is the same on both S1 and S3, and I suspect that even the PCB for the hashing boards is the same, meaning I could buy bare chips from bitmaintech and replace the 1st gen ones, with the new ones.
At least, that's what I would have done if I had to produce miners, using the exact same pinout on each chips generation to be able to reuse the "old" boards I have in stock.
would also cost less, since ou don't have to design a new board.

Control board IS the same, however the PCB boards are definitely different. The S1 has 32 chips on each PCB whereas the S3 has only 16 chips. However, if the chip footprints and connections are the same, then you can swap these out from the S1, but you'll need more power to run the hungrier S3 chips (now that you'll have 32) AND you'll also have to swap out the voltage regulators AND hope that the power traces on the PCB can handle the extra power.

All in all, I think there's still a bit to hack on this S1, both hardware wise and driver / software wise. Also the miner. There are still LOTS of efficiencies that I believe can be squeezed out of the old lady.

Ok, then, just populate 4 to 6 chips in each row, wire the 2 in/out pins from the missing chips to bypass them, and/or run the S3 chips undervolted.

Yep, lots of permutations, like I said, the old lady still has some juice left in her!
Did you use one of these?

powersup

sr. member

Activity: 406

Merit: 250

Quote from: pekatete on September 12, 2014, 08:03:12 AM

Quote from: J4bberwock on September 12, 2014, 07:56:06 AM

Quote from: powersup on September 12, 2014, 07:54:18 AM

ok, official legend status goes to J4bberwock

It took me almost 1 hour Grin

YOU are the man. I think it is safe to say the bounty goes to you, and you deserve it.
For RPi I think the same solution works out of the box, simply install the cgminer version and you are off.
TOP MAN!

I couldn't agree more. J4bberwock deserves the bounty. I have sent 0.025 BTC to the address in your signature. Tx id = e13405cc62b73c17b4290c24fd59b0f99d9a14523f98e678716cc37320c40953.

J4bberwock

hero member

Activity: 840

Merit: 1000

Quote from: pekatete on September 12, 2014, 08:14:15 AM

Quote from: J4bberwock on September 12, 2014, 08:06:14 AM

The control board is the same on both S1 and S3, and I suspect that even the PCB for the hashing boards is the same, meaning I could buy bare chips from bitmaintech and replace the 1st gen ones, with the new ones.
At least, that's what I would have done if I had to produce miners, using the exact same pinout on each chips generation to be able to reuse the "old" boards I have in stock.
would also cost less, since ou don't have to design a new board.

Control board IS the same, however the PCB boards are definitely different. The S1 has 32 chips on each PCB whereas the S3 has only 16 chips. However, if the chip footprints and connections are the same, then you can swap these out from the S1, but you'll need more power to run the hungrier S3 chips (now that you'll have 32) AND you'll also have to swap out the voltage regulators AND hope that the power traces on the PCB can handle the extra power.

All in all, I think there's still a bit to hack on this S1, both hardware wise and driver / software wise. Also the miner. There are still LOTS of efficiencies that I believe can be squeezed out of the old lady.

Ok, then, just populate 4 to 6 chips in each row, wire the 2 in/out pins from the missing chips to bypass them, and/or run the S3 chips undervolted.

pekatete

hero member

Activity: 518

Merit: 500

Quote from: J4bberwock on September 12, 2014, 08:06:14 AM

The control board is the same on both S1 and S3, and I suspect that even the PCB for the hashing boards is the same, meaning I could buy bare chips from bitmaintech and replace the 1st gen ones, with the new ones.
At least, that's what I would have done if I had to produce miners, using the exact same pinout on each chips generation to be able to reuse the "old" boards I have in stock.
would also cost less, since ou don't have to design a new board.

Control board IS the same, however the PCB boards are definitely different. The S1 has 32 chips on each PCB whereas the S3 has only 16 chips. However, if the chip footprints and connections are the same, then you can swap these out from the S1, but you'll need more power to run the hungrier S3 chips (now that you'll have 32) AND you'll also have to swap out the voltage regulators AND hope that the power traces on the PCB can handle the extra power.

All in all, I think there's still a bit to hack on this S1, both hardware wise and driver / software wise. Also the miner. There are still LOTS of efficiencies that I believe can be squeezed out of the old lady.

J4bberwock

hero member

Activity: 840

Merit: 1000

Quote from: pekatete on September 12, 2014, 08:03:12 AM

Quote from: J4bberwock on September 12, 2014, 07:56:06 AM

Quote from: powersup on September 12, 2014, 07:54:18 AM

ok, official legend status goes to J4bberwock

It took me almost 1 hour Grin

YOU are the man. I think it is safe to say the bounty goes to you, and you deserve it.
For RPi I think the same solution works out of the box, simply install the cgminer version and you are off.
TOP MAN!

yes, it's easier to use a simple uart to usb board, maybe with a usb hub chip.
It will allow to plug it on any controler, PC, mac, RPI, openwrt platform...

Topic: [DIY] - Reward $100 | Antminer S1/S3 Blade on Raspberry Pi - page 40. (Read 82179 times)