[GUIDE] BitFury Miner Support/Tuning

[juhakall]

Two quick questions on the goxxed info.  1) What points are you using exactly to measure the altered voltage?   I am leery to probe around the board while powered up for fear of shorting something.
                                                        2) What exactly is reniceing and why is it important to reduce?  This is the first time I have seen this term used in the bitfury threads.

Thanks very much and awesome work!  

I'm really curious about 1) too.

Reniceing is just the linux way to set processing priority for a task. You might want the miner process to have a high priority. I'd expect this to not matter much, or be set by default. Maybe it really helps, I'm trying on mine now.

[darkfriend77]

Two quick questions on the goxxed info.  1) What points are you using exactly to measure the altered voltage?   I am leery to probe around the board while powered up for fear of shorting something.
                                                        2) What exactly is reniceing and why is it important to reduce?  This is the first time I have seen this term used in the bitfury threads.

Thanks very much and awesome work!  

I'm really curious about 1) too.

Reniceing is just the linux way to set processing priority for a task. You might want the miner process to have a high priority. I'd expect this to not matter much, or be set by default. Maybe it really helps, I'm trying on mine now.

:-) jep, maybee a small picture ... where and how to ... measure the altered voltage would help ... us ... .-)

[-Redacted-]

"Nice" is a term used to describe the priority of a process running under UNIX. The nicer a task is, the more it "yields" to others and the lower the priority that it runs at.  "Renice" - means changing the nice value, typically increasing the nice value and lowering the priority of a task.  Since interrupts are pre-emptive, tasks don't really "yield" - something with a higher nice value gets preempted (interrupted) more often.

The root user can lower the nice value of a process, increasing its priority and lengthing the interval between preempts - making it "less nice".  It is possible to lower a task's nice value to the point where it has more priority than the kernel, thereby guaranteeing that your Unix system will crash if your code makes a system call of any kind.  On a system that is not very busy, raising or lowering the nice value doesn't have much of an effect.    

[Cablez]

Thanks for the info on niceing guys.  As you may have guessed I am not a big 'nix guy....learning.  Looks like I do this on windows a lot for demanding processes so its just a matter of vocabulary. 

Has there been info given yet about increasing the priority of the mining process yielding better hash results?

[dacman61]

This pencil mod trick has been working great! At first I used too much pencil and had a resistance of 1.08k and the noncerate shot up to 35Gh/s then dropped down quickly... Soon after I wiped away some of the pencil marking, I finally read a resistance of 1.18k, and this turned out to be the sweet spot for me... I've been steady at 30.5-32Gh/s noncerate now. I also put a small turbo desk fan blowing on it at low speed.

Code:

16      AIfDSo  55      1.904   2.315   133     7       0       0       219     [0:F]   436     9 9 8 9 9 7 8 8 8 8 8 8 9 9 9 7         0 0 1 0 0 2 1 1 0 0 0 0 0 0 0 2
speed:880 noncerate[GH/s]:32.098 (2.006/chip) hashrate[GH/s]:34.542 good:2242 errors:190 spi-err:0 miso-err:0 jobs:375 cores:42% good:16 bad:0 off:0 (best[GH/s]:31.525) Sun Sep 15 13:38:13 2013
0:      880     32.098  34.542  2242    190     0       0       16      0       0       (2.006/chip)    42%

Code:

16      AIfDSo  55      1.847   2.304   129     14      0       0       218     [0:F]   363     9 8 7 8 7 8 8 8 8 9 8 8 9 7 9 8         0 1 2 1 2 1 1 0 1 0 1 1 0 2 0 1
speed:880 noncerate[GH/s]:30.594 (1.912/chip) hashrate[GH/s]:34.584 good:2137 errors:176 spi-err:0 miso-err:0 jobs:375 cores:51% good:16 bad:0 off:0 (best[GH/s]:32.098) Sun Sep 15 13:43:13 2013
0:      880     30.594  34.584  2137    176     0       0       16      0       0       (1.912/chip)    51%

I've been running now for the past 30 minutes at the time of me gathering this info.

How does this look to you?

[juhakall]

This pencil mod trick has been working great! At first I used too much pencil and had a resistance of 1.08k and the noncerate shot up to 35Gh/s then dropped down quickly... Soon after I wiped away some of the pencil marking, I finally read a resistance of 1.18k, and this turned out to be the sweet spot for me... I've been steady at 30.5-32Gh/s noncerate now. I also put a small turbo desk fan blowing on it at low speed.

Code:

16      AIfDSo  55      1.904   2.315   133     7       0       0       219     [0:F]   436     9 9 8 9 9 7 8 8 8 8 8 8 9 9 9 7         0 0 1 0 0 2 1 1 0 0 0 0 0 0 0 2
speed:880 noncerate[GH/s]:32.098 (2.006/chip) hashrate[GH/s]:34.542 good:2242 errors:190 spi-err:0 miso-err:0 jobs:375 cores:42% good:16 bad:0 off:0 (best[GH/s]:31.525) Sun Sep 15 13:38:13 2013
0:      880     32.098  34.542  2242    190     0       0       16      0       0       (2.006/chip)    42%

Code:

16      AIfDSo  55      1.847   2.304   129     14      0       0       218     [0:F]   363     9 8 7 8 7 8 8 8 8 9 8 8 9 7 9 8         0 1 2 1 2 1 1 0 1 0 1 1 0 2 0 1
speed:880 noncerate[GH/s]:30.594 (1.912/chip) hashrate[GH/s]:34.584 good:2137 errors:176 spi-err:0 miso-err:0 jobs:375 cores:51% good:16 bad:0 off:0 (best[GH/s]:32.098) Sun Sep 15 13:43:13 2013
0:      880     30.594  34.584  2137    176     0       0       16      0       0       (1.912/chip)    51%

I've been running now for the past 30 minutes at the time of me gathering this info.

How does this look to you?

Great result, though it sounds weird that the hashrate wasn't stable at first with 1.08k. BTW, how do you get these stats: "cores:51% good:16 bad:0 off:0"? I just have the record (it's called 'record' on my stats, not 'best') right after jobs. Are those some extra stats that are shown when autotune is enabled?

[dacman61]

I'm running this command as suggested by -Redacted-:

Code:

while (true) do tail -3 /run/shm/.stat.log; sleep 305; done

Latest reading:

Code:

16      AIfDSo  55      1.933   2.325   135     19      0       0       220     [0:F]   197     8 9 10 8 9 9 9 8 9 8 8 8 7 8 8 9        2 1 0 2 1 1 1 2 0 1 1 1 2 1 2 1
speed:880 noncerate[GH/s]:32.270 (2.017/chip) hashrate[GH/s]:34.753 good:2254 errors:165 spi-err:0 miso-err:0 jobs:295 cores:71% good:16 bad:0 off:0 (best[GH/s]:32.255) Sun Sep 15 13:58:13 2013
0:      880     32.270  34.753  2254    165     0       0       16      0       0       (2.017/chip)    71%

[GH]

Two quick questions on the goxxed info.  1) What points are you using exactly to measure the altered voltage?   I am leery to probe around the board while powered up for fear of shorting something.
                                                        2) What exactly is reniceing and why is it important to reduce?  This is the first time I have seen this term used in the bitfury threads.

Thanks very much and awesome work! 

Meanwhile I have a stable nonce rate of average >35 Gh/s on both boards - so time to give something back:

You can measure the voltage on nearly every capacitor on the board, I usually use the bigger ones directly beside the DC/DC converter, left to the coil (the bigest part on the board). There are not many ways to do something wrong, just make sure to not shorten your probes, as there are 30+ Amps available. (I have exactly 0.840 Volts there with 55er clocks and it looks very stable, although there is no easy way to measure the real current draw on the core voltage and confirm that everything is inside bounds.)

BTW: I also reported about a mysterious chip, which usually throws lots of errors, but sometimes recovers completely after a longer uptime. When running this board alone or in switched order with the other one, the errors of this single chip also disappear.

[Foofighter]

Two quick questions on the goxxed info.  1) What points are you using exactly to measure the altered voltage?   I am leery to probe around the board while powered up for fear of shorting something.
                                                        2) What exactly is reniceing and why is it important to reduce?  This is the first time I have seen this term used in the bitfury threads.

Thanks very much and awesome work! 

Meanwhile I have a stable nonce rate of average >35 Gh/s on both boards - so time to give something back:

You can measure the voltage on nearly every capacitor on the board, I usually use the bigger ones directly beside the DC/DC converter, left to the coil (the bigest part on the board). There are not many ways to do something wrong, just make sure to not shorten your probes, as there are 30+ Amps available. (I have exactly 0.840 Volts there with 55er clocks and it looks very stable, although there is no easy way to measure the real current draw on the core voltage and confirm that everything is inside bounds.)

BTW: I also reported about a mysterious chip, which usually throws lots of errors, but sometimes recovers completely after a longer uptime. When running this board alone or in switched order with the other one, the errors of this single chip also disappear.

Hi, thanks for sharing.

Can you also confirm the Ohm values given above?

[GH]

Not really, because I do not like quick'n'dirty stuff like the pencil mod. I have a real resistance of 4.07k with 1% metal film (not measured mounted inside the circuit), so you can't compare easily. I suppose this is one reason why it is so stable, regardless of temperature. (Carbon an graphite have a negative temperature coefficient, which causes the resistance to drop with rising temperatures, which leads to even more heat and so on...)

I'm runnning 5 "server grade" fans on the 2 boards, which make a hell of a noise, but keep the boards and surfaces of the components pretty cool.

Edit: Added NTC explaination, typos, wrong value

[juhakall]

I did the pencil mod on 2 boards, which maxed out at 47 GH/s prior to modding. Both boards' R02F resistors were modified from 1.3k to 1.18k ohm. Now the noncerate is 59-61 GH/s. Here are the results before and after modding. As you can see, chip #10 stopped making errors. Also, chip #24 stopped requiring a 24 hour period after a reboot until it stops making errors. I added a fan and the highest temps I could measure from the boards dropped from ~80C to ~60C. I'll report if anything changes.

[joris]

...

g) I Use only one pool for mining

h) renice the miner proxy process to -14 or lower

Code:

 2100 root        6 -14 33196 18864  4064 R 81.0  3.8 36h27:30 python ./mining_proxy.py -o us3.eclipsemc.com -p 3333 -cu XXXXXXXXXXX -cp XXXX -gp 8332 -v

h.1) I use htop to do it
h.1.1) sudo apt-get install htop
h.1.2) pi@bitfury / $ sudo htop
h.1.3) Press F7 repeatedly to reduce nice.
        

This did a lot for my performance: boost of 5 GH on a starter kit (regard as highly anecdotical, single measurement). But it seems to have a logic: instead of 3 processes consuming 50-60% of CPU time, now only 1 'large' process exists, using around 20% of the CPU at a high priority. It seems to relieve the raspi a lot...

Does it make sense to only prioritise the mining_proxy.py? I see many more chainminer processes in htop...

[zurg]

Ok, modded my starter kit.

1.307 on both boards to 1.215 and 1.229(EOL)
Went from around
1: 24.234GH/s
2: 14.379GH/s
to
1: 29.234GH/s
2: 16.779GH/s

Bitminter now reporting from 42 to 49GH/s (was nearly 50 a sec ago)
from about ~35 to 39 before.

I am in a server room at 62F with fan blowing over. Chips are lukewarm.

Definitely at least a 5Gh/s increase!

[spiccioli]

Hi,

I've found, because of an error, that sometimes a bad chip can hash at low speed using "aIFdso" instead of the normal "aIfDSo" flag string.

Code:

18	aIFdso	47	0.687	0.888	48	26	0	0	84	[1:1]	288	4 3 3 4 1 2 3 2 2 3 4 2 4 4 3 4 	1 2 1 0 3 2 1 2 3 2 1 3 1 1 2 1 
47	aIFdso	40	0.272	0.835	19	26	0	0	79	[2:E]	443	1 2 1 1 2 1 1 1 2 2 0 1 1 1 1 1 	2 1 2 1 0 1 2 2 1 1 3 2 2 2 2 2 
72	aIFdso	47	0.487	0.930	34	19	0	0	88	[4:7]	288	2 3 2 2 2 1 3 2 1 3 2 3 2 2 2 2 	1 0 1 1 1 2 0 2 3 1 2 1 1 1 1 1 
97	aIFdso	47	0.515	0.814	36	24	0	0	77	[6:0]	288	1 1 2 2 2 1 2 3 2 3 3 2 4 1 3 4 	3 2 1 1 1 3 2 1 2 1 1 2 0 3 1 0 
143	aIFdso	40	0.344	0.909	24	22	0	0	86	[8:E]	504	2 2 1 2 2 2 1 2 0 2 1 1 1 2 0 3 	1 1 2 1 1 1 2 1 3 1 2 1 1 1 3 0 
151	aIFdso	47	0.458	0.814	32	14	0	0	77	[9:6]	288	2 2 2 3 3 2 2 3 2 1 1 1 2 2 3 1 	1 1 1 0 0 1 1 0 1 1 1 2 1 1 0 2 
162	aIFdso	47	0.558	0.814	39	17	0	0	77	[A:1]	288	2 3 1 4 4 1 2 2 3 3 3 3 1 1 3 3 	2 1 3 0 0 2 1 1 0 0 0 0 2 3 1 1 
178	aIFdso	40	0.329	0.824	23	20	0	0	78	[B:1]	552	3 0 2 2 1 1 1 1 0 3 1 2 1 2 2 1 	0 3 1 0 1 1 1 1 3 0 2 1 2 1 1 2 
182	aIFdso	40	0.086	0.835	6	45	0	0	79	[B:5]	656	1 1 0 0 0 1 0 0 0 2 0 0 1 0 0 0 	2 2 3 3 4 3 4 3 3 1 3 3 2 3 3 3 
200	aIFdso	48	0.716	0.877	50	0	0	0	83	[C:7]	36	3 3 3 3 3 3 3 3 3 3 3 3 4 4 3 3 	0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
205	aIFdso	44	0.544	0.698	38	0	0	0	66	[C:C]	36	2 3 3 3 3 3 3 2 2 2 2 2 2 2 2 2 	0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
225	aIFdso	40	0.501	0.835	35	19	0	0	79	[E:0]	288	3 3 1 2 1 3 2 2 0 3 3 2 3 3 2 2 	1 1 2 1 2 0 1 1 3 0 0 1 1 1 2 2 
238	aIFdso	40	0.415	0.708	29	15	0	0	67	[E:D]	289	1 3 1 2 2 2 3 2 2 3 1 2 2 0 1 2 	2 0 2 1 1 1 0 1 1 0 1 0 0 2 2 1 
254	aIFdso	40	0.229	0.846	16	42	0	0	80	[F:D]	596	0 0 0 1 3 2 1 0 2 1 1 1 1 1 1 1 	4 4 4 3 1 1 2 3 1 2 2 3 3 3 3 3 
256	aIFdso	44	0.415	0.877	29	14	0	0	83	[F:F]	324	2 1 2 1 3 3 3 1 2 2 1 2 1 2 1 2 	1 2 1 2 0 0 0 2 0 0 1 0 1 1 2 1 

What does it really change using "Internal Fast" instead of "Internal Divided Slow" ?

I've made tests lowering speed down to 10 without conclusive findings, btw boards are extremely "instable", simply restarting miner can be enough for a good chip to start throwing errors that slow it down/shut it down.

I'm going to change miner so that best.conf speed is the minimum it can go irregardless of errors of any kind while keeping the auto-tune flag active.

I'm doing this to see if I can stop the spiralling down of the unit which tends to slow down if left mining alone with auto-tune on.

spiccioli

ps. note that I've changed my miner to accept speeds lower than 52 and that I've not pencil-modded them (yet).

[EvilLizardApparel]

Using pencil trick, I had an H-Card go from 21 GH/s to 39 GH/s . However, after 60 seconds the hashrate slowly decreases to lower than 21 GH/s. Is this a result of the auto-tuning?

EDIT: When miner is restarted, card goes back to 39 GH/s .. but then does the same thing.

[goxed]

Using pencil trick, I had an H-Card go from 21 GH/s to 39 GH/s . However, after 60 seconds the hashrate slowly decreases to lower than 21 GH/s. Is this a result of the auto-tuning?

EDIT: When miner is restarted, card goes back to 39 GH/s .. but then does the same thing.

My guess-
The regulator autoshutting

[btc4life]

I just added some heatsinks to one of my cards today to test them out before I committed to buying enough for all 16 cards.

They seem to fit the surface area of each chip perfectly.

http://www.amazon.com/Cosmos-Aluminum-Cooling-Heatsinks-cooler/dp/B007XACV8O/ref=sr_1_1?s=electronics&ie=UTF8&qid=1379368738&sr=1-1&keywords=Cosmos+%C2%AE+20+PCS+mini+Aluminum+Chips

http://i106.photobucket.com/albums/m264/atunicus/DSCN0803.jpg

[zurg]

Using pencil trick, I had an H-Card go from 21 GH/s to 39 GH/s . However, after 60 seconds the hashrate slowly decreases to lower than 21 GH/s. Is this a result of the auto-tuning?

EDIT: When miner is restarted, card goes back to 39 GH/s .. but then does the same thing.

What  did you set it to?
Mine has been rock solid all day. Few rare errors, like 150 out of over 200k.

[-Redacted-]

I just added some heatsinks to one of my cards today to test them out before I committed to buying enough for all 16 cards.

They seem to fit the surface area of each chip perfectly.

http://www.amazon.com/Cosmos-Aluminum-Cooling-Heatsinks-cooler/dp/B007XACV8O/ref=sr_1_1?s=electronics&ie=UTF8&qid=1379368738&sr=1-1&keywords=Cosmos+%C2%AE+20+PCS+mini+Aluminum+Chips

I thought these chips were designed to put heat out through the bottom, and that heatsinks should go on the back of the board over the thermal vias...

[EvilLizardApparel]

Using pencil trick, I had an H-Card go from 21 GH/s to 39 GH/s . However, after 60 seconds the hashrate slowly decreases to lower than 21 GH/s. Is this a result of the auto-tuning?

EDIT: When miner is restarted, card goes back to 39 GH/s .. but then does the same thing.

My guess-
The regulator autoshutting

....Is there a way to run it and not have it autoshutdown? It was hashing like a boss...Heat was fine...is it the auto-tune, or a hardware limit I'm hitting?