Pages:
Author

Topic: [ANN] Bitfury ASIC sales in EU and Europe - page 72. (Read 250482 times)

legendary
Activity: 1792
Merit: 1047
September 05, 2013, 12:12:09 PM
I had a similar issue ... somehow if I use load balanced mining .... with 40-42GH on 3 pools im getting 13GH at BTCGuild ... 13GH .... at eligius ... and 12 GH at Bitminter ...

if I change to mine 100% on BTCGuild it shows me 20-30 GH only .... i tryed with 16,32 & 64 diff

Have you tried the reported accepted shares from BTCGuild to the accepted shares the client reports?

This way you can compare with accuracy.
sr. member
Activity: 434
Merit: 265
September 05, 2013, 11:52:25 AM
I had a similar issue ... somehow if I use load balanced mining .... with 40-42GH on 3 pools im getting 13GH at BTCGuild ... 13GH .... at eligius ... and 12 GH at Bitminter ...

if I change to mine 100% on BTCGuild it shows me 20-30 GH only .... i tryed with 16,32 & 64 diff
legendary
Activity: 1225
Merit: 1000
September 05, 2013, 11:45:39 AM
Sorry to interrupt you spiccioli.

My kit is showing me a Noncerate of 43.1GH/s (yayy! a gigahash more compared to autotuning)

But on BTCGuild I'm only at 33.3GH/s (99.32% valid shares)

I know the guild only shows an estimate, but the rate has always been considerably lower than what the raspbi tells me over the last 30 hours of uninterrupted mining.

Is that regular behaviour?
sr. member
Activity: 251
Merit: 250
September 05, 2013, 11:28:59 AM
I agree, the solder jumper will surely cause problems.
hero member
Activity: 574
Merit: 501
September 05, 2013, 11:27:34 AM
Good eye, Spicciolli!  That looks like a problem, for certain...  Fix that (or get it fixed), and you've probably repaired the card.
legendary
Activity: 1379
Merit: 1003
nec sine labore
September 05, 2013, 11:12:05 AM
Half of the chips, 7-16, are dead and go to zero with auto-tuning while not hashing with manual tuning.
Since the SPI signals form a long chain from chip to chip, there's probably a problem at chip 6 or 7. Check those extra carefully on the board. Note: the chips are numbered in white ink in the same order as in the software.

If the chips are mounted correctly, and one is broken, it could be that it prevents communication with next one in the chain. In that case, the chip could be removed, and the solder jumpers shorted. Beware: don't attempt to fix unless you've done stuff like this before.

Ok, I'll look at them more in depth.

I've never used a solder iron in my life, so I won't try to break it more than it already is Smiley

spiccioli

I've made a couple more photos, to my untrained eye the only strange thing is SJ51

http://imgur.com/c8hCggO,A4ImRve

Chips 6,7,8,9 seem to have their IO pins OK.

spiccioli
legendary
Activity: 1379
Merit: 1003
nec sine labore
September 05, 2013, 10:33:07 AM
Half of the chips, 7-16, are dead and go to zero with auto-tuning while not hashing with manual tuning.
Since the SPI signals form a long chain from chip to chip, there's probably a problem at chip 6 or 7. Check those extra carefully on the board. Note: the chips are numbered in white ink in the same order as in the software.

If the chips are mounted correctly, and one is broken, it could be that it prevents communication with next one in the chain. In that case, the chip could be removed, and the solder jumpers shorted. Beware: don't attempt to fix unless you've done stuff like this before.

Ok, I'll look at them more in depth.

I've never used a solder iron in my life, so I won't try to break it more than it already is Smiley

spiccioli
sr. member
Activity: 251
Merit: 250
September 05, 2013, 10:28:46 AM
Half of the chips, 7-16, are dead and go to zero with auto-tuning while not hashing with manual tuning.
Since the SPI signals form a long chain from chip to chip, there's probably a problem at chip 6 or 7. Check those extra carefully on the board. Note: the chips are numbered in white ink in the same order as in the software.

If the chips are mounted correctly, and one is broken, it could be that it prevents communication with next one in the chain. In that case, the chip could be removed, and the solder jumpers shorted. Beware: don't attempt to fix unless you've done stuff like this before.
sr. member
Activity: 251
Merit: 250
September 05, 2013, 10:24:27 AM
Chip 4F has a blob of semi-transparent material on one side, on I/O pins
That's flux residue from somebody fixing a short with soldering iron.

hero member
Activity: 574
Merit: 501
September 05, 2013, 10:22:24 AM
The semi-transparent material is flux, which you might want to clean off - see up a message or two.  Shorted Vdd pins are not an issue as long as they are shorted to each other and not other pins.  Use a magnifying glass, look for solder bridges on both the front and back of the board.
legendary
Activity: 1379
Merit: 1003
nec sine labore
September 05, 2013, 10:19:01 AM
This is my worst performing card,

http://imgur.com/YMdcFdt,54T7aWB,8iOjWKl,Cp0EGFi

Chip 4F has a blob of semi-transparent material on one side, on I/O pins, nearly half of the chips have shorted VDD pins.

Half of the chips, 7-16, are dead and go to zero with auto-tuning while not hashing with manual tuning.

Any idea?

spiccioli
member
Activity: 84
Merit: 10
September 05, 2013, 08:23:24 AM
Update on H-card soldering issues.
I sent some H-cards to be examined and reworked, and most of them came out clean. The bridges seem to be caused by flux resin in over 95% of cases.

Should we follow this instructions: https://nuxx.net/wiki/Flux_Removal to clean all boards?
hero member
Activity: 525
Merit: 500
..yeah
September 05, 2013, 08:10:59 AM
Update on H-card soldering issues.
I sent some H-cards to be examined and reworked, and most of them came out clean. The bridges seem to be caused by flux resin in over 95% of cases.
That sounds promising, any chances one can get a dead chip caused by this error to work again by some simple process?
hero member
Activity: 560
Merit: 500
September 05, 2013, 08:10:36 AM
Some info on setup

Chainminer is compiled with encoded username and password that are hard coded in jobconnect.cpp. Currently miner is set to run three getwork threads to localhost 127.0.0.1:8332-8334.

Code:
hosts_t hosts[]={
{"Basic dHl0dXMucGkyOnB1YmxpY3Bhc3M=","http://127.0.0.1:8332/",{0,0,0,0,0,0,0,0xFFFFFFFF},NULL,NULL,0,0,0} // local stratum client
,{"Basic dHl0dXMucGkyOnB1YmxpY3Bhc3M=","http://127.0.0.1:8333/",{0,0,0,0,0,0,0,0xFFFFFFFF},NULL,NULL,0,0,0} // local stratum client
,{"Basic dHl0dXMucGkyOnB1YmxpY3Bhc3M=","http://127.0.0.1:8334/",{0,0,0,0,0,0,0,0xFFFFFFFF},NULL,NULL,0,0,0} // local stratum client
};

If you want to use the miner without stratum proxy you may do so by setting it in the above place. However, you must first encode your miner credentials to be inserted there:

Code:
pi@bitfury01 /opt/bitfury/chainminer $ echo -n username:password | base64
dXNlcm5hbWU6cGFzc3dvcmQ=
pi@bitfury01 /opt/bitfury/chainminer $

Then you must copy these encoded credentials into jobconnect.cpp and rebuild the miner:

Code:

dXNlcm5hbWU6cGFzc3dvcmQ=

hosts_t hosts[]={
{"Basic dXNlcm5hbWU6cGFzc3dvcmQ=","http://getworkpool:port/",{0,0,0,0,0,0,0,0xFFFFFFFF},NULL,NULL,0,0,0} //
};


Save the file and make. Restart miner.
member
Activity: 84
Merit: 10
September 05, 2013, 07:56:21 AM
Are you sure? I'd say from my limited testing that stopping/starting the miner is enough for it to load a new /opt/bitfury/best.cnf

spiccioli
I can confirm this, with fairly extensive testing.

Yep, seems to be working. My poor Pi, it has been rebooted several times for no reason Sad

After rebooting best.cnf was copied into /run/shm. It looks like for first time after modifying /opt/bitfury/best.cnf you must reboot the device, and after that only miner process needs to be stopped and start again.

Thanks a lot!
hero member
Activity: 560
Merit: 500
September 05, 2013, 07:50:47 AM
Update on H-card soldering issues.
I sent some H-cards to be examined and reworked, and most of them came out clean. The bridges seem to be caused by flux resin in over 95% of cases.
hero member
Activity: 910
Merit: 1000
Items flashing here available at btctrinkets.com
September 05, 2013, 07:49:36 AM
I wanted to post something motivational for a change:

sr. member
Activity: 466
Merit: 250
September 05, 2013, 07:47:20 AM
Are you sure? I'd say from my limited testing that stopping/starting the miner is enough for it to load a new /opt/bitfury/best.cnf

spiccioli
I can confirm this, with fairly extensive testing.

Yep, seems to be working. My poor Pi, it has been rebooted several times for no reason Sad
hero member
Activity: 910
Merit: 1000
Items flashing here available at btctrinkets.com
September 05, 2013, 06:58:21 AM
Are you sure? I'd say from my limited testing that stopping/starting the miner is enough for it to load a new /opt/bitfury/best.cnf

spiccioli
I can confirm this, with fairly extensive testing.
legendary
Activity: 1379
Merit: 1003
nec sine labore
September 05, 2013, 06:56:48 AM
My Kit has best performance after few minutes from start, before autotune starts messing with chip settings. After that everything falls down. Will it be ok to copy /run/shm/.stat.log to /opt/bitfury/best.cnf ? Or those files have different structure?

Also earlier when I have tried to modify settings I have fallen into problem. I have problem with using best.cnf for starting the miner. I made a copy of .best.log to best.cnf in /opt/bitfury directory, after that I have switched autotune off on few chips in .cnf file and saved the changes.
Then I have started the miner with console output in screen. After several minutes of mining if i open .stat.log all chips I have set to not use autotune, are set to autotune on, and have different speeds then those in /opt/bitfury/best.cnf file. Also there is no copy of best.cnf in /run/shm directory after miner starts. Am I suppose to copy best.cnf into /run/shm manually before starting the miner?

.stat.log and best.cnf have the structure. Keep in mind that .stat.log is changing. If I were you, I would copy the .best.log.

Restarting the miner does not work. You need to reboot the Pi (sudo reboot)

Are you sure? I'd say from my limited testing that stopping/starting the miner is enough for it to load a new /opt/bitfury/best.cnf

spiccioli
Pages:
Jump to: