Hacking The KNC Firmware: Overclocking

[elasticband]

clock setting 291, been running fr over a week now

[1713558969]

1713558969

[1713558969]

1713558969

[1713558969]

1713558969

[1713558969]

1713558969

[tiozes]

clock setting 271, been running fr over a week now

As voltage has to die each orient?

[elasticband]

I used the method of flashing 99.2E then back to 99.2, this enables the 8vrm's and then i just proceeded to tune and OC. Been running very stable with low HW and some buzzing noise for about a week.

the volts are pretty even among each pair, nothing additional done, no problems here.

[tiozes]

I used the method of flashing 99.2E then back to 99.2, this enables the 8vrm's and then i just proceeded to tune and OC. Been running very stable with low HW and some buzzing noise for about a week.

the volts are pretty even among each pair, nothing additional done, no problems here.

ok, so my jupiter vrm October 4, 8vrms could change, I am not very clear about this.

[tolip_wen]

Edit to say, after to put again the firm 0.99.1 tune, i have a little problem with voltages.. it was crazy.. now is working well after reset to factory defaults in advanced but a line 12v of my 1100w PSU now is dead, i got a screenshot, any idea of this?

The VRM are designed for 40W, many of us regularly push them to between 50-60W.
At 70W per VRM I think you are on your own in new territory!

Your screenshot shows 70W per die (per VRM also).
70W x 4 die = 280W per ASIC

Add AT LEAST 10% to that because the VRM's use that much before even cooking them by overclocking.

OVER 300 Watt per ASIC is what your power supply was asked to do before failure.
(you were asking for >1200W from a 1100W supply)

I suggest 300W per ASIC is possibly too much.
Almost definately too much for a single PCIe 6 pin power connector.
You can install another on each module, KnCMiner provided for that on the PCB.

YMMV

[tolip_wen]

@ bondus

The whole Rtune feedback loop has 3 flavors with just the Oct. units.

8 VRM units have 4 output caps for each VRM located right next to the ASIC on top.
4 VRM units have 6 output caps, same 4 plus 2 around the corner of the ASIC.
('same' as in location, value is undetermined)
4 on plus 4 off have 8 output caps per active VRM +/- whatever the idle VRM contributes.

Calculating Rtune depends on # of caps and on an 8 VRM unit it changes.

From the keyboard side what to do?
Would artificially altering the gain change the response time for similar effect?
Could this be tolerated or compensated for?

YMMV

[idee2013]

Hi

How it's possible to make script that copies my own cgminer.sh script to /etc/init.d/cgminer.sh on startup?

And how to make script that restarts the miner with like 3 hour intervals? As there has been some downtime with overclocking and unit just shutting down=)

now 500gh/s semi-stable 2 module november saturn, was on during the night. Started clocking each die individually. To make it stable i needed to get rid of those fault 4:s turning up after some time. I think they have timer on them and that causes some overtemp-protection to trigger in after some time.

Values for asic 4 are 361 and asic 0 varies in between 361 and 2C1.


Description=cgminer 3.9.0|SUMMARY

MHS av=503541.30
MHS 5s=503582.25

Work Utility=7034.48
Difficulty Accepted=245831.00000000
Difficulty Rejected=1248.00000000
Difficulty Stale=0.00000000
Best Share=376240
Device Hardware%=0.1790
Device Rejected%=0.5065
Pool Rejected%=0.5051
Pool Stale%=0.0000|

ASIC Board    Info
0    

Temperature sensor: 46.0 C
Die ID    Cores ON    Cores OFF    %
0   48   0    100
1   48   0    100
2   48   0    100
3   48   0    100
DC/DC ID    ON/OFF    Status    Input Voltage    Output Voltage    Output Current
0    OFF    OK    11.5 V    0.82 V    35.2 A (28.9 W)
1    OFF    OK    11.4 V    0.822 V    34.6 A (28.4 W)
2      OFF    OK    11.3 V    0.815 V    37.5 A (30.6 W)
3    OFF    OK    11.4 V    0.817 V    37.7 A (30.8 W)
4    OFF    OK    11.3 V    0.821 V    35.9 A (29.5 W)
5    OFF    OK    11.4 V    0.818 V    36.1 A (29.5 W)
6    OFF    OK    11.3 V    0.826 V    29.5 A (24.4 W)
7     OFF    OK    11.3 V    0.82 V    30.2 A (24.8 W)
4    

Temperature sensor: 47.5 C
Die ID    Cores ON    Cores OFF    %
0   48   0    100
1   48   0    100
2   48   0    100
3   48   0    100
DC/DC ID    ON/OFF    Status    Input Voltage    Output Voltage    Output Current
0    OFF    OK    11.5 V    0.82 V    38.2 A (31.3 W)
1    OFF    OK    11.4 V    0.821 V    37.8 A (31 W)
2    OFF    OK    11.5 V    0.822 V    37.4 A (30.7 W)
3    OFF    OK    11.5 V    0.817 V    37.6 A (30.7 W)
4    OFF    OK    11.5 V    0.821 V    37.5 A (30.8 W)
5    OFF    OK    11.4 V    0.816 V    37.8 A (30.8 W)
6    OFF    OK    11.5 V    0.816 V    36.6 A (29.9 W)
7    OFF    OK    11.4 V    0.817 V    36.6 A (29.9 W)

Total DC/DC power output: 471 W

vrm's for asic 0 die 3 make fault 4 risen above that 30.  Asic 0 is heating more than 4 in overall.

has anyone tried this with an oct junit with 8 activted VRMS? i wonder why everyone with  with 8 vrms boards oct reaching only 790Gh/s max.

[crashoveride54902]

Hi

How it's possible to make script that copies my own cgminer.sh script to /etc/init.d/cgminer.sh on startup?

And how to make script that restarts the miner with like 3 hour intervals? As there has been some downtime with overclocking and unit just shutting down=)

now 500gh/s semi-stable 2 module november saturn, was on during the night. Started clocking each die individually. To make it stable i needed to get rid of those fault 4:s turning up after some time. I think they have timer on them and that causes some overtemp-protection to trigger in after some time.

Values for asic 4 are 361 and asic 0 varies in between 361 and 2C1.


Description=cgminer 3.9.0|SUMMARY


Total DC/DC power output: 471 W

vrm's for asic 0 die 3 make fault 4 risen above that 30.  Asic 0 is heating more than 4 in overall.

has anyone tried this with an oct junit with 8 activted VRMS? i wonder why everyone with  with 8 vrms boards oct reaching only 790Gh/s max.

because oct. jups just ain't made to oc this much...

[tiozes]

Edit to say, after to put again the firm 0.99.1 tune, i have a little problem with voltages.. it was crazy.. now is working well after reset to factory defaults in advanced but a line 12v of my 1100w PSU now is dead, i got a screenshot, any idea of this?

The VRM are designed for 40W, many of us regularly push them to between 50-60W.
At 70W per VRM I think you are on your own in new territory!

Your screenshot shows 70W per die (per VRM also).
70W x 4 die = 280W per ASIC

Add AT LEAST 10% to that because the VRM's use that much before even cooking them by overclocking.

OVER 300 Watt per ASIC is what your power supply was asked to do before failure.
(you were asking for >1200W from a 1100W supply)

I suggest 300W per ASIC is possibly too much.
Almost definately too much for a single PCIe 6 pin power connector.
You can install another on each module, KnCMiner provided for that on the PCB.

YMMV

The power supply was broken, leaving to run a 12v rail.

Now he's running a cooler master 1300w, and it seems that everything works normally.

have if I can get more power, which would secure V for each die since I have high temperatures.

[elasticband]

is there instructions on permanent writing the clock settings?

[elasticband]

any ideas why i get such bad results with 5 boards, which i have done with 4 boards.

It is almost as if when i delete the cgminer and i add the overclock setting  from my text file, it stops the boards from accepthing the OC settings.

Should i be changing certain variables in my text file for the added 5th board?

or are control boards just not able to handle OC settings with 5-6 boards?


thanks

[rmarleal]

I lost the cgminer.ssh file to do the overclocking after upgrading to the new firmware 1.00 and going back to 0.99.1 Any help here please?

this is what i can see when i use putty

https://i.imgur.com/xY8xBJQ.jpg

[Aggrophobia]

anyone who can upload modified cgminer.sh-backup?

[mekadeka]

You know, in hindsight and upon reflecting on it, I find it concerning that KNC sold different equipment under the same model name. That is real problematic and really potentially shady to me.

[tyrion70]

anyone who can upload modified cgminer.sh-backup?

http://pastebin.com/xS4U9DXC <-- this ones at 211

[bkminer]

Hello, has anyone done anything with the 1.00 update?  Maybe tell me how to get the 1000Mhz setting for an October Jupiter...

Thanks,

[elasticband]

any ideas of what variables to change to make 5 boards take the OC settings.

edit: I have tried adding 4 to the end of 0 1 2 3 at the 2 or 3 points i could find this mentioned bellow, but it did not have an effect.

eidt: if anyone can edit the bellow for me so it works with 5 boards i would be extremely grateful and would offer a small reward

#!/bin/sh

PATH=/usr/local/sbin:/usr/local/bin:/sbin:/bin:/usr/sbin:/usr/bin

use_bfgminer=
if [ -f /config/miner.conf ]; then
        . /config/miner.conf  #if bfgminer is enabled, this file contains: use_bfgminer="true"
fi
if [ "$use_bfgminer" = true ] ; then
        DAEMON=/usr/bin/bfgminer
        NAME=bfgminer
        DESC="BFGMiner daemon"
        EXTRA_OPT="-S knc:auto"
else
   DAEMON=/usr/bin/cgminer
   NAME=cgminer
   DESC="Cgminer daemon"
   EXTRA_OPT=
fi


set -e

test -x "$DAEMON" || exit 0

do_start() {
   # Stop SPI poller
   spi_ena=0
   i2cset -y 2 0x71 2 $spi_ena

   good_ports=""
   bad_ports=""

   # CLear faults in megadlynx's
   for b in 3 4 5 6 7 8 ; do
      for d in 0 1 2 3 4 5 6 7 ; do
         i2cset -y $b 0x1$d 3 >/dev/null 2>&1 || true
      done
   done

   for p in 0 1 2 3 4 5 ; do
      i2cset -y 2 0x71 1 $((p+1))
      good_flag=0
      ar="$(spi-test -s 50000 -OHC -D /dev/spidev1.0 0x80,3,0,0,0,0,0,0 | tail -c 13)"
                if [ "x$ar" = "x00 30 A0 01" ] ; then
         good_flag=1
      fi
      ar="$(spi-test -s 50000 -OHC -D /dev/spidev1.0 0x80,2,0,0,0,0,0,0 | tail -c 13)"
                if [ "x$ar" = "x00 30 A0 01" ] ; then
         good_flag=1
      fi
      ar="$(spi-test -s 50000 -OHC -D /dev/spidev1.0 0x80,1,0,0,0,0,0,0 | tail -c 13)"
                if [ "x$ar" = "x00 30 A0 01" ] ; then
         good_flag=1
      fi
      ar="$(spi-test -s 50000 -OHC -D /dev/spidev1.0 0x80,0,0,0,0,0,0,0 | tail -c 13)"
                if [ "x$ar" = "x00 30 A0 01" ] ; then
         good_flag=1
      fi

      if [ "$good_flag" = "1" ] ; then
         good_ports=$good_ports" $p"
      else
         bad_ports=$bad_ports" $p"
      fi
   done

   if [ -n "$good_ports" ] ; then
      for p in $good_ports ; do
         # Re-enable PLL
         i2cset -y 2 0x71 1 $((p+1))
         for c in 0 1 2 3 ; do
            cmd=$(printf "0x84,0x%02X,0,0" $c)
            spi-test -s 50000 -OHC -D /dev/spidev1.0 $cmd >/dev/null
            cmd=$(printf "0x86,0x%02X,0x02,0x91" $c)
            spi-test -s 50000 -OHC -D /dev/spidev1.0 $cmd >/dev/null
            cmd=$(printf "0x85,0x%02X,0,0" $c)
            spi-test -s 50000 -OHC -D /dev/spidev1.0 $cmd >/dev/null
         done
         
         # re-enable all cores
         i=0
         while [[ $i -lt 192 ]] ; do
            i2cset -y 2 0x2$p $i 1
            i=$((i+1))
         done
         spi_ena=$(( spi_ena | (1 << $p) ))
      done
   fi
   if [ -n "$bad_ports" ] ; then
      for p in $bad_ports ; do
         # Disable PLL
         i2cset -y 2 0x71 1 $((p+1))
         for c in 0 1 2 3 ; do
            cmd=$(printf "0x84,0x%02X,0,0" $c)
            spi-test -s 50000 -OHC -D /dev/spidev1.0 $cmd >/dev/null
         done
         
         # disable all cores
         i=0
         while [[ $i -lt 192 ]] ; do
            i2cset -y 2 0x2$p $i 0
            i=$((i+1))
         done
         spi_ena=$(( spi_ena & ~(1 << $p) ))
      done
   fi

   # Disable direct SPI
   i2cset -y 2 0x71 1 0

   # Enable SPI poller
   i2cset -y 2 0x71 2 $spi_ena

   start-stop-daemon -b -S -x screen -- -S cgminer -t cgminer -m -d "$DAEMON" --api-listen -c /config/cgminer.conf $EXTRA_OPT
}

do_stop() {
        killall -9 bfgminer cgminer 2>/dev/null || true
}
case "$1" in
  start)
        echo -n "Starting $DESC: "
   do_start
        echo "$NAME."
        ;;
  stop)
        echo -n "Stopping $DESC: "
   do_stop
        echo "$NAME."
        ;;
  restart|force-reload)
        echo -n "Restarting $DESC: "
        do_stop
        do_start
        echo "$NAME."
        ;;
  *)
        N=/etc/init.d/$NAME
        echo "Usage: $N {start|stop|restart|force-reload}" >&2
        exit 1
        ;;
esac

exit 0

[tiozes]

The Bermod 0.4 operates in the firm 1.0? Júpiter october 4vrm

[j03]

Edit to say, after to put again the firm 0.99.1 tune, i have a little problem with voltages.. it was crazy.. now is working well after reset to factory defaults in advanced but a line 12v of my 1100w PSU now is dead, i got a screenshot, any idea of this?

The VRM are designed for 40W, many of us regularly push them to between 50-60W.
At 70W per VRM I think you are on your own in new territory!

Your screenshot shows 70W per die (per VRM also).
70W x 4 die = 280W per ASIC

Add AT LEAST 10% to that because the VRM's use that much before even cooking them by overclocking.

OVER 300 Watt per ASIC is what your power supply was asked to do before failure.
(you were asking for >1200W from a 1100W supply)

I suggest 300W per ASIC is possibly too much.
Almost definately too much for a single PCIe 6 pin power connector.
You can install another on each module, KnCMiner provided for that on the PCB.

YMMV

Good advice! Greed is NOT good Mr. Gecko!

[crashoveride54902]

any ideas of what variables to change to make 5 boards take the OC settings.

edit: I have tried adding 4 to the end of 0 1 2 3 at the 2 or 3 points i could find this mentioned bellow, but it did not have an effect.

eidt: if anyone can edit the bellow for me so it works with 5 boards i would be extremely grateful and would offer a small reward

#!/bin/sh

PATH=/usr/local/sbin:/usr/local/bin:/sbin:/bin:/usr/sbin:/usr/bin

use_bfgminer=
if [ -f /config/miner.conf ]; then
        . /config/miner.conf  #if bfgminer is enabled, this file contains: use_bfgminer="true"
fi
if [ "$use_bfgminer" = true ] ; then
 

exit 0

you can try this....http://pastebin.com/AsGZKu4E

no clue if it'll work, i just used the individual die setting to try n oc all your mods...i've set'em all at 291 for ya...Let me know if it worked

The Bermod 0.4 operates in the firm 1.0? Júpiter october 4vrm

yes its working for me on 1.0fw