Topic: [GUIDE] BitFury Miner Support/Tuning - page 9. (Read 148062 times)

The "pencil mod" is done on the H-cards. Most people are measuring voltage from the top of the inductor on the H-card to the GND screw on the M-board. IMO, it's much better to use the individual H-card's ground though, because there's a significant difference that increases the further the card is from the power connections on the M-board. Using the M-board's GND screw, you might get a measurement of 0.90v at the farthest card and 0.85v at the nearest, while the true voltage on the cards is 0.83v for both. The chips only care about the voltage difference between the power and ground planes on their own card, not the voltage rise on the ground bus across the M-board. A good ground point on an H-card is the exposed ground pad on the back side of the card, behind the regulator IC.

So I got the new image installed and nothing comes up when I get into bfgminer. Is there some argument I am missing starting up bfgminer?

I was under the impression that pencil mods were done on h-boards, not the m-board?  I do have heatsinks on every card, back of both chips and regulators as well so that should be covered.

got it working guys. turns out that BTCGuild was having problems with the server kicking off the bitfury. btcguild fix that problems now.

bitfury changed the exact chip voltage a few times between batches. most batches have some extra headroom to pencil mod the voltages higher but there are a few batches that almost benefit from the opposite.

take a multimeter and check the voltage beween the top contact on the inductor and a GND terminal on the m-board. 0.8-0.88V is the ideal range if you have reasonable amount of heatsinks AND airflow. If its <0.8V, do a pencil mod. if its >0.88V, improve the cooling. add a small heatsink to the thermal vias of the regulator on the backside of the board. this chip does a lot of work and needs just as much cooling as the ASICs

Any thermal paste with metal can be electrically capacitive, but are not conductive. This means while the thermal paste may hold a charge, it does not transfer a charge from one object to another. But their is always a chance of something going wrong if you just slather it on there.

Sorry if I'm being annoying, but could someone explain how the dab of thermal paste doesn't short the chip connections on the back of the H-board?

I'm new to electrical engineering, and don't want to make a mistake.

EDIT: So far, this is the best explanation I've found. Could someone explain if it's correct? Thanks.

If the pool is reporting significantly lower hashrate than the miner, restart the proxies and/or chainminer.

If cards are acting up even at low speeds, power off, reseat all cards, power back on. Alternatively, without powering off, you could just rock the cards side to side and front to back to loosen any oxidation on the contacts without fully pulling them out of the rig. Restart chainminer after this to re-init all chips.

Restart proxies and/or chainminer if all chips are reporting 0GH.

this is my experience as well along with several others.

I have gone almost a month at 400+gh and now recently my reported speeds are more like 200-300 at the pool.  Reported rates on the unit itself are in the 300-350 range but they used to be almost 500.  I have applied heatsinks to the backs of the chips and voltage regulater area and this is a climate controlled room with cool air blowing over the cards.  This is the new version hardware as well.  I regularly have cards going to zero, I manually tune them as far as 49 and they work at low rates and another card goes to zero, always at least one card down, sometimes as many as three.  I am using a 1200w power supply as well so it's hard to believe it's because they aren't getting juice. 

Any thoughts?

Welcome to BitFury.

After I ran the stop-stratumproxy.sh, start-stratumproxy.sh, stop-miner.sh and start-miner.sh scripts it started working. No idea what was wrong.

Create /opt/bitfury/best.cnf like this to run all chips at speed 52 with autotune disabled or to whatever you desire:

Thank you for taking the time to reply.

Couple things to add - Ive had this up and running since December. Ive had the speeds up to 55 (880) but AC kept turning off, so I lowered down to 52 until I got that fixed. The AC was fixed today so I bumped the speeds up to 55 again and proceeded with the reboot.

Logically followed it would imply that I screwed up the best.cnf file or something. This happened once before where I had did a search and replace all of 55 to 52, but forgot to go back and change the chip numbers 55, 155 and 255. Once I fixed this the miner started up fine. Knowing I had done this once before I manually checked all the chip numbers and they are sequentially correct.

Is there any way to make/restore a default best.cnf?

make sure you wait at least a few min for hashing to occur.  make sure you have enough cooling via fans +/- heatsinks.  try rearranging some h boards.  other than that, rebooting should've worked.

Can someone running a full rig with chainminer and proxy, please provide some info about load average and CPU idle time?

My 400GH/s Bitfury rig is not hashing after sudo reboot command

Output of /run/shm/.stat.log after reboot:

1       AIfDSo  52      0.000   0.000   0       0       0       0       0      $
2       AIfDSo  52      0.000   0.000   0       0       0       0       0      $
3       AIfDSo  52      0.000   0.000   0       0       0       0       0      $

Output of /run/shm/.stat.log after a few minutes following reboot:

1       AiFDso  52      0.000   0.000   0       0       0       0       0      $
2       AiFDso  52      0.000   0.000   0       0       0       0       0      $
3       AiFDso  52      0.000   0.000   0       0       0       0       0      $

As you can see, the autotune is turning off the chips.

Please help!

=========
extra output

pi@0-6-249:~$ sudo /opt/bitfury/start-miner-console.sh
SPEED: min:52 def:52 max:53
M-Board version 2 detected
INIT: 256 chips detected

=========

speed:13312 noncerate[GH/s]:0.000 (0.000/chip) hashrate[GH/s]:0.000 good:0 errors:0 spi-err:0 miso-err:0 duplicates:0 jobs:0 cores:0% good:256 bad:0 off:0 (best[GH/s]:$
board-2 speed   nrate   hrate   good    errors  spi-err miso-er duplic  good    bad     off     per chip        good cores
0:      832     0.000   0.000   0       0       0       0       0       16      0       0       (0.000/chip)    0%      speed down
1:      832     0.000   0.000   0       0       0       0       0       16      0       0       (0.000/chip)    0%
2:      832     0.000   0.000   0       0       0       0       0       16      0       0       (0.000/chip)    0%
3:      832     0.000   0.000   0       0       0       0       0       16      0       0       (0.000/chip)    0%
4:      832     0.000   0.000   0       0       0       0       0       16      0       0       (0.000/chip)    0%
5:      832     0.000   0.000   0       0       0       0       0       16      0       0       (0.000/chip)    0%
6:      832     0.000   0.000   0       0       0       0       0       16      0       0       (0.000/chip)    0%
7:      832     0.000   0.000   0       0       0       0       0       16      0       0       (0.000/chip)    0%
8:      832     0.000   0.000   0       0       0       0       0       16      0       0       (0.000/chip)    0%
9:      832     0.000   0.000   0       0       0       0       0       16      0       0       (0.000/chip)    0%
A:      832     0.000   0.000   0       0       0       0       0       16      0       0       (0.000/chip)    0%
B:      832     0.000   0.000   0       0       0       0       0       16      0       0       (0.000/chip)    0%
C:      832     0.000   0.000   0       0       0       0       0       16      0       0       (0.000/chip)    0%
D:      832     0.000   0.000   0       0       0       0       0       16      0       0       (0.000/chip)    0%
E:      832     0.000   0.000   0       0       0       0       0       16      0       0       (0.000/chip)    0%
F:      832     0.000   0.000   0       0       0       0       0       16      0       0       (0.000/chip)    0%


====================

make sure you change the pool settings as they currently are pointed at his pool 

Awesome. I wasn't sure about that one, but I will give it a try. Thanks for the help.

i'm successfully running salfter's image of bfgminer from here:

https://bitcointalksearch.org/topic/m.4511347

just installed it yesterday and the great news is it seems to have solved my "reset" problems i described earlier.  that was my first priority in evaluating bfgminer.  i still need to fine tune them as the hashing rate is not optimal with a fair # of HW errors on some rigs but i'm very much encouraged.