Pages:
Author

Topic: Antminer D3 Blissz firmware (10/12 v1.12 update) - page 32. (Read 125923 times)

newbie
Activity: 30
Merit: 0
v2.02 beta :: 14/12 :: test version

Download:

https://mega.nz/#!DghG2KwC!OJEK4_d47skRB6yrLRE2bfzlX7FWXLXzgoKA_q5dKco


Changes 14/12:
- fixed auto recover from xxxx
- cleaned up kernel logging
- a bit more aggressive tuning
- tunes down faster now when too many HW errors are detected


Changes 13/12:
- Totally rewritten code base, based from latest available bitmain / cgminer sources
- ASIC's now automatically recovers from xxxxx errors without restarting (The auto reboot options are still in there in case it doesn't succeed)
- blazing fast mining status overview in the webinterface Cheesy
- as requested: the dev mining starts later now (after 5 minutes)
- as requested: dev pools status is now visible in the miner status

and last but not least...

- added a first version of the auto tune!  Cool

When enabled this will search for the best possible frequency at the given voltage setting. The first 5 to 10 minutes it will search for a good base to start from. After that it checks for excessive HW errors every 15 minutes and adjusts the frequency accordingly.

Important: This feature is not completely finished and the startup is a bit slow now.. Therefore I need your help. So what do you need to do:

set your advanced settings the same as displayed bellow
let it run for at least 20 minutes
PM your (complete) kernel log
By doing so I can get to a conclusion what the starting frequency of the auto tune algorithm should be.
In case you want to help more: increase the voltage a bit and do the steps above again as it will give some more information

Thank you!

https://img15.hostingpics.net/pics/453940Untitled.jpg

Please test and give feedback / a kernel log how it works for you. I need some feedback on this version guys  Grin


My test is up there.
sr. member
Activity: 336
Merit: 258
Hello,
to be on the safe side is it possible to add something like this.

if temp > 100 degrees ; killall -9 cgminer

I will soon test version 2.

Will make it a bit more advanced as sometimes one chip can give a bad reading, but something like that is useful yes
newbie
Activity: 33
Merit: 0
Hello,
to be on the safe side is it possible to add something like this.

if temp > 100 degrees ; killall -9 cgminer

I will soon test version 2.
member
Activity: 130
Merit: 10
what version is that?
i tried your 1.12 and gives HW every minute or so, at lowest voltage and 400 freq.
sr. member
Activity: 336
Merit: 258
v2.02 beta :: 14/12 :: test version

Download:

https://mega.nz/#!DghG2KwC!OJEK4_d47skRB6yrLRE2bfzlX7FWXLXzgoKA_q5dKco


Changes 14/12:
- fixed auto recover from xxxx
- cleaned up kernel logging
- a bit more aggressive tuning
- tunes down faster now when too many HW errors are detected


Changes 13/12:
- Totally rewritten code base, based from latest available bitmain / cgminer sources
- ASIC's now automatically recovers from xxxxx errors without restarting (The auto reboot options are still in there in case it doesn't succeed)
- blazing fast mining status overview in the webinterface Cheesy
- as requested: the dev mining starts later now (after 5 minutes)
- as requested: dev pools status is now visible in the miner status

and last but not least...

- added a first version of the auto tune!  Cool

When enabled this will search for the best possible frequency at the given voltage setting. The first 5 to 10 minutes it will search for a good base to start from. After that it checks for excessive HW errors every 15 minutes and adjusts the frequency accordingly.

Important: This feature is not completely finished and the startup is a bit slow now.. Therefore I need your help. So what do you need to do:

set your advanced settings the same as displayed bellow
let it run for at least 20 minutes
PM your (complete) kernel log
By doing so I can get to a conclusion what the starting frequency of the auto tune algorithm should be.
In case you want to help more: increase the voltage a bit and do the steps above again as it will give some more information

Thank you!



Please test and give feedback / a kernel log how it works for you. I need some feedback on this version guys  Grin
sr. member
Activity: 336
Merit: 258


---

Dec 14 04:59:15 (none) local0.err cgminer[376]: low hashrate detection trigger: 31.464000 Gh/s
Dec 14 04:59:15 (none) local0.notice cgminer[376]: set voltage = 1143.942031  real:255 mv
Dec 14 04:59:15 (none) local0.notice cgminer[376]: set_PIC16F1704_voltage_new ok, voltage = 0xff
Dec 14 04:59:16 (none) local0.notice cgminer[376]: set_PIC16F1704_voltage_new ok, voltage = 0xff
Dec 14 04:59:16 (none) local0.notice cgminer[376]: set_PIC16F1704_voltage_new ok, voltage = 0xff
Dec 14 04:59:16 (none) local0.err cgminer[376]: cgminer time error total_secs = 1513227556.664247 last_total_secs = 1.000000
Dec 14 04:59:17 (none) local0.err cgminer[376]: D1 0: work 7f not find error



i see 31,4GH/s at @1150W  that its true? its constant?

No this is the low hashrate trigger for triggering the reboot when needed, but I see it's not correct during startup as it's still tuning the frequency.
newbie
Activity: 12
Merit: 0


---

Dec 14 04:59:15 (none) local0.err cgminer[376]: low hashrate detection trigger: 31.464000 Gh/s
Dec 14 04:59:15 (none) local0.notice cgminer[376]: set voltage = 1143.942031  real:255 mv
Dec 14 04:59:15 (none) local0.notice cgminer[376]: set_PIC16F1704_voltage_new ok, voltage = 0xff
Dec 14 04:59:16 (none) local0.notice cgminer[376]: set_PIC16F1704_voltage_new ok, voltage = 0xff
Dec 14 04:59:16 (none) local0.notice cgminer[376]: set_PIC16F1704_voltage_new ok, voltage = 0xff
Dec 14 04:59:16 (none) local0.err cgminer[376]: cgminer time error total_secs = 1513227556.664247 last_total_secs = 1.000000
Dec 14 04:59:17 (none) local0.err cgminer[376]: D1 0: work 7f not find error



i see 31,4GH/s at @1150W  that its true? its constant?
newbie
Activity: 30
Merit: 0
KERNEL LOG: https://ghostbin.com/paste/7t8fh

My # CHAIN1 card does not work, so the log should show it faulty.

Today BitMain support me answered this here:

"Dear Marcos,

Thanks for your consultation!

Please try to replace with a good known control board. Kindly let me know if it works.

This is a control board.
https://shop.bitmain.com/productDetail.htm?pid=00020170712085151361v8GpNlSS06B7

Please let us know if you have additional questions or concerns.

Best regards,
Jocelyn
Bitmain"


And now, what should I do, Blissz?
newbie
Activity: 12
Merit: 0
Thank you, but I think that we can go with lowest voltage to 500mhz and in that case power consumption is more than 1100w, I tried this 2 weak ago, dont know if something is change in meantime with new fw..
i doont try more than 425Mhz because at 425Mhz the HW rise to much on my D3. Maybe i try another high frecuenci to see if is jusg because my D3 dont like 425Mhz. (Same one say samething bout machine dont like specific frecuency hehehe let give him same trust xD)
full member
Activity: 303
Merit: 105
Thank you, but I think that we can go with lowest voltage to 500mhz and in that case power consumption is more than 1100w, I tried this 2 weak ago, dont know if something is change in meantime with new fw..
newbie
Activity: 12
Merit: 0
Thank You Blissz for your custom firmware. version tested v1.12 from 10 december.
I want to mention this custom firmware is good for undervolt option the overclock i dont know i dont try it.
i want to mention the efficency are the same as stock firmware or maybe slightr better the Blissz firmware.
Because on stock firmware my D3 run at 1050W = 19Ghz  with Blissz frimware(lowest voltage option) run with 610-630W = ~15Ghz
And i think this is good for hot summer or if noise is a problem.
i want to say my power supply is APW5(2600W on 220v) from Bitmain.
And i think is a litell bit unestable ... on antpool its like mountain up and down the hashrate(see picture).
i let you some picture.
https://imgur.com/ZxFg4Cj
https://imgur.com/wi4D3So
https://imgur.com/CNlssAr
https://imgur.com/2kuM1KT
https://imgur.com/orbauDU
https://imgur.com/o6bEtM1


PD: This are the value of your question micairvas see picture.
If lowest voltage option is set on autotune, how far cpu frequency can go and what is max power consumption on lowest voltage?
full member
Activity: 303
Merit: 105
If lowest voltage option is set on autotune, how far cpu frequency can go and what is max power consumption on lowest voltage?
jr. member
Activity: 169
Merit: 1
i do not have a problem to autotune for 2-3hours, better safe then sorry:)
also my questino is, if tunning is chain one by one or all together? let me explain what i mean:
i set voltage stage 1 frequency 400. it will start to tune it with first chain only, looking for the sweet pot (based on fan speed percentage or auto it depend and temp temperature for chips (70-75 maximum?))
then it will do the same for chain two and then chain 3.
jr. member
Activity: 137
Merit: 2
hi blissz, you asked for feedback and tips.

a good start for autotune is a 'lastgood' variable, containing after every tune the last working level of mh/s each chain. So you could set the new starting value after reboot to 'lastgood' - 18 as example...

How it works now: It does goes up in speed and remembers the frequencies that gave little to no HW errors. After it found a maximum (3 times consequently too many HW errors) it steps back a few steps and then a slower algorithm kicks in and checks each chain for HW errors. If the HW errors are above 6 in 5 minutes it will step back one frequency (in the list of good frequencies)

Any suggestions are appreciated.
some people say, change freq must be only once at 10 minutes.
bitmaintech recommend wait 20 minutes for testing before send blade for repair.

Else, my bad experiments show it, your D3's blades will may be ruined after some time (many HW for short time, always).

ps: they will be self-healed, may be, after resetting to factory settings and one-two day waiting (with worked device). but may be not. my D3 is healed, this way. but few HW still present anyway after this.
sr. member
Activity: 336
Merit: 258
sr. member
Activity: 336
Merit: 258
hi blissz, you asked for feedback and tips.

a good start for autotune is a 'lastgood' variable, containing after every tune the last working level of mh/s each chain. So you could set the new starting value after reboot to 'lastgood' - 18 as example...

How it works now: It does goes up in speed and remembers the frequencies that gave little to no HW errors. After it found a maximum (3 times consequently too many HW errors) it steps back a few steps and then a slower algorithm kicks in and checks each chain for HW errors. If the HW errors are above 6 in 5 minutes it will step back one frequency (in the list of good frequencies)

Any suggestions are appreciated.
newbie
Activity: 32
Merit: 0
u need to have in mind that this is a BETA its still for testing and needs further improvements  Roll Eyes
and pls dont post the logs here  Cheesy
sr. member
Activity: 336
Merit: 258
I'm getting tons of HW and rejects with the auto tune. It needs more refinement.

A new version is on the way, but please PM your kernel log. The autotune is also quite slow yet, give it some time ( 30 minutes at least)
newbie
Activity: 33
Merit: 0
I'm getting tons of HW and rejects with the auto tune. It needs more refinement.
member
Activity: 301
Merit: 10
Hey Blissz, sometimes my hashboards are disappeared after i changed something in the setting, i have to reboot my miners to make them detect my hashboards again.

Also, one of my hashboards shows 1 HW, and then that "1 HW" disappeared for a while, is this normal with your latest FW?
Pages:
Jump to: