Hacking The KNC Firmware: Overclocking

[crashoveride54902]

leaving everything at default and just changing to 211 your rig shuts down but does not with the default?

exactly

If so, try your Jupiter with 3 modules plugged in, if that works then you are just at the edge of your power supply.

Its a 1050W PSU so i dont expect it to be at the limit, yet. I suspect that the Seasonic is very sensitive when it comes to little power peaks and shuts itself down.

What PSU are known to work with those overclocked jupiters?

I'm using a 1000w seasonic plat. psu using 231 so mabye your psu is going out or can't handle the extra strain on 12v rail...you could also try taking out the pcie extender cables, I noticed a drop in input volts when i had those hooked up...have another psu to test it out? I haven't had an issue with mine for oc'in yet...

[1714043874]

1714043874

[1714043874]

1714043874

[1714043874]

1714043874

[ImI]

leaving everything at default and just changing to 211 your rig shuts down but does not with the default?

exactly

If so, try your Jupiter with 3 modules plugged in, if that works then you are just at the edge of your power supply.

Its a 1050W PSU so i dont expect it to be at the limit, yet. I suspect that the Seasonic is very sensitive when it comes to little power peaks and shuts itself down.

What PSU are known to work with those overclocked jupiters?

I'm using a 1000w seasonic plat. psu using 231 so mabye your psu is going out or can't handle the extra strain on 12v rail...you could also try taking out the pcie extender cables, I noticed a drop in input volts when i had those hooked up...have another psu to test it out? I haven't had an issue with mine for oc'in yet...

yes, i will try other PSUs as soon as i will be back at the hosting place (i dont have the jupiters at home)
i have heard of taking out the pci cables but i am not sure if i got that straight. so you suggest to unplug the blue cables that knc delivered and connect the boards directly to the psu cable? are those long enough to reach all four modules with the PSU?

[thorvald]

I cant paste any images here, i see, avascript (void) says the link for "insert image".

Somebody asked for "advanced" tab images, but i don't have any advanced tab on my miner page?! It shows only voltages for vrms (asic 0, voltages are 0,825

STATUS=S
When=1389894564
Code=11
Msg=Summary
Description=cgminer 3.9.0|SUMMARY
Elapsed=2433
MHS av=370243.17
MHS 5s=361734.32
Found Blocks=0
Getworks=79
Accepted=569
Rejected=0
Hardware Errors=78
Utility=14.03
Discarded=154
Stale=0
Get Failures=0
Local Work=215334
Remote Failures=0
Network Blocks=4
Total MH=900701886.6114
Work Utility=5172.80
Difficulty Accepted=199150.00000000
Difficulty Rejected=0.00000000
Difficulty Stale=0.00000000
Best Share=186928
Device Hardware%=0.0372
Device Rejected%=0.0000
Pool Rejected%=0.0000
Pool Stale%=0.0000|

ASIC Board    Info
0    

Temperature sensor: 57.0 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    11.7 V    0.825 V    25.3 A (20.9 W)
1    OFF    OK    11.6 V    0.828 V    26.1 A (21.6 W)
2    OFF    OK    11.6 V    0.823 V    26.5 A (21.8 W)
3    OFF    OK    11.6 V    0.824 V    26.3 A (21.7 W)
4    OFF    OK    11.6 V    0.828 V    26.8 A (22.2 W)
5    OFF    OK    11.6 V    0.826 V    27.2 A (22.5 W)
6    OFF    OK    11.5 V    0.828 V    26.7 A (22.1 W)
7    OFF    OK    11.6 V    0.824 V    27.1 A (22.3 W)

4    

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    11.7 V    0.829 V    27 A (22.4 W)
1    OFF    OK    11.6 V    0.828 V    26.9 A (22.3 W)
2    OFF    OK    11.6 V    0.828 V    26.8 A (22.2 W)
3    OFF    OK    11.6 V    0.824 V    27.3 A (22.5 W)
4    OFF    OK    11.7 V    0.829 V    27.6 A (22.9 W)
5    OFF    OK    11.6 V    0.826 V    26.7 A (22.1 W)
6    OFF    OK    11.7 V    0.825 V    26.4 A (21.8 W)
7    OFF    OK    11.7 V    0.826 V    25.8 A (21.3 W)

Total DC/DC power output: 351 W

the dump has been taken like after 15-20min after starting cgminer 3.9.

I started now cgminer3.9 with modified cgminer.sh ( 0x86 setting 265 again ). It seems to stable but there might be tendency for to go down. Average on cgminer shows 369. I want to see if poolside shows bigger rate than with 3.8.5. With it it was 365, but on the kncminer stats page it said 374gh/s.

Changin the last digit to 6 kills everything, vrms just go to error and after that only 2 were on status "ok", all other with fault xx or off.

I wish there was more info that what these registers really are and how they affect things.

is this a November device ?
How did you activated all the dc/dc converters for the October unit ?

[temen]

Hi its november device.

VRMs are running quite hot, can't keep finger on them at all. This and possible future overclockin in mind I got 4 old graphics cards with small aluminium coolers from recycling center. Cost 10 euros. (They didn't accept bitcoin! ). Sawed these in 4 pieces and got 16 small coolers. Just bonded them with arctic silver adhesive (too expensive).

I quickly checked from the pdf that 0x80 is master clock and 0x86 is divider, if this pdf is related to Perhaps I'll try fidling with the master clock tonight.. Not so sure It always scares the shit out of me when amps to berzerk and everything just.. stops

EDIT: don't know where i did get that 0x80 is master clock, I don't think it is

[complang]

Seems to work pretty well with 265 on nov saturn cgminer 3.9.0. Total output according to bertmod = 351W (+20W from OEM). 25-28A current/VRM.
Hashrate varying between 350-370 but it's for sure a speed increse
So far the temps seems to stabilize around 46c with 0,15% HW. Amazing!

What does the "spi-test" do really? Running it at 50000 right now.

Edit/update: Avg hashing rate stabilizes at 360Gh/s. + 20 000 shares on BTC Guild (+~12-15%)

Stats:
STATUS=S
When=1390166000
Code=11
Msg=Summary
Description=cgminer 3.9.0|SUMMARY
Elapsed=179569
MHS av=360975.24
MHS 5s=361370.11
Found Blocks=0
Getworks=6576
Accepted=60400
Rejected=407
Hardware Errors=28968
Utility=20.18
Discarded=12597
Stale=0
Get Failures=0
Local Work=15590432
Remote Failures=1
Network Blocks=362
Total MH=64820127467.5121
Work Utility=5042.77
Difficulty Accepted=15155066.00000000
Difficulty Rejected=62756.00000000
Difficulty Stale=0.00000000
Best Share=8287478
Device Hardware%=0.1916
Device Rejected%=0.4158
Pool Rejected%=0.4124
Pool Stale%=0.0000|

[CeeCee]

Hi,

i just started to fiddle with my 4VRM jupiter.

I tuned to 2.5V SPI and 256k SPIFrequenzy first and then started with 201.

It fires up, Amps go to around 50 and Temps around 60C, and then the miner suddenly shuts down.

Its most probable a PSU shutdown (i am using a Seasonic 1050 Gold) cause after some minutes it boots just normal again.

Any ideas?

dont change anything under the advanced tab. set everything to default and try 211. it is working very good for me.

i did and again miner shuts down after ≈ 10sec

what PSU do you use?

Corsair AX1200

[tiozes]

hi, for my jupiter 4vrm October with OC to 201 with 211 going well and gives less performance that has 201? when would have to give more power.
the firm is 0.99.1tune.

regards

[r1senfa17h]

I now have my 6-module Oct Jupiter running stable @ 1026GH/s  . The key was temperature for me. I lose cores left and right with another 5 degrees Celsius. I placed a box fan blowing air downward and my HW errors dropped to < 0.5% and all cores stay active.

[ImI]

I now have my 6-module Oct Jupiter running stable @ 1026GH/s  .

at what amps?

[r1senfa17h]

I now have my 6-module Oct Jupiter running stable @ 1026GH/s  .

at what amps?

58-61.5

[DPoS]

I now have my 6-module Oct Jupiter running stable @ 1026GH/s  .

at what amps?

58-61.5

that's some serious clockin..   reminds me of the first week we got them and all the heat issues

[idee2013]

I now have my 6-module Oct Jupiter running stable @ 1026GH/s  .

at what amps?

58-61.5

that's some serious clockin..   reminds me of the first week we got them and all the heat issues

i have avg 12h at eligius pool ~680Gh/s with only 50-55amps. 4 asics oct jup and 231 setting.

[ImI]

I now have my 6-module Oct Jupiter running stable @ 1026GH/s  .

at what amps?

58-61.5

that's some serious clockin..   reminds me of the first week we got them and all the heat issues

i have avg 12h at eligius pool ~680Gh/s with only 50-55amps. 4 asics oct jup and 231 setting.

extra cooling?

[tiozes]

hi, for my jupiter 4vrm October with OC to 201 with 211 going well and gives less performance that has 201? when would have to give more power.
the firm is 0.99.1tune.

regards

help, because 231 have much less power than a 201? as I can get to 670-680GH?

October 4 asics jup

help please

[idee2013]

hi, for my jupiter 4vrm October with OC to 201 with 211 going well and gives less performance that has 201? when would have to give more power.
the firm is 0.99.1tune.

regards

help, because 231 have much less power than a 201? as I can get to 670-680GH?

October 4 asics jup

help please

hi,

https://bitcointalk.org/index.php?topic=313978.msg4425943#msg4425943

have 99.1 tune, too

4vrm asic dies -0,1003 volts each
each VRM got  VRAM heatsinks and i have a big fan over the Jupiter

[CeeCee]

HELL YEA! After 2 hours I was able to overclock my 8VRM unit with 211. SPI 3.3V 256kHZ and now ne magic behind this. You just have to find a good value for each DIE to get ~50A.

1. edit cgminer.sh with '211'
2. restart cgminer '/etc/init.d/cgminer.sh restart'
3. give it a minute and have a look at the advanced tab.
4. write down wich DIEs don't reach the 50A.
5. give each die more volt. 0.0748 seems to be a good starting point. but to be sure go up at lower steps.
6. now restart again cgminer. only at this point you will see if you where successful.
7. did you reached 50A/DIE? | no? -> restart from point 2. | yes? have fun!

i am not responsible for anything

[padrino]

HELL YEA! After 2 hours I was able to overclock my 8VRM unit with 211. SPI 3.3V 256kHZ and now ne magic behind this. You just have to find a good value for each DIE to get ~50A.

1. edit cgminer.sh with '211'
2. restart cgminer '/etc/init.d/cgminer.sh restart'
3. give it a minute and have a look at the advanced tab.
4. write down wich DIEs don't reach the 50A.
5. give each die more volt. 0.0748 seems to be a good starting point. but to be sure go up at lower steps.
6. now restart again cgminer. only at this point you will see if you where successful.
7. did you reached 50A/DIE? | no? -> restart from point 2. | yes? have fun!

i am not responsible for anything

This definitely shows the general process however there is a lot of variability in the units.. For example a number of them can do 211 without any voltage changes, others like yours require a fairly heavy set of modifications. The point is to take an easy and not just jump to the top voltage wise..

[elasticband]

every machine requires different tweaking, other peoples results give a base line of where to start trying though.


8vrm for me = 3.3spi 1755428mhz. ranges from 0.07 - 0.1 volts each Die. ranges above 50a and 40w

Currently getting 640GH/s 650w (without cntrl board) and 0.91% HW .

4vrm is completely different.

[temen]

No advanced tuning, some cooling done better:


Mining Status
CGMiner Status   Running (pid=4739)
Last Checked   Mon Jan 20 14:57:46 UTC 2014
Avg. Hash Rate   459 Gh/s
WU   6418
Difficulty Accepted   152907
HW Status
ASIC slot #1   46.0 ℃
ASIC slot #2   -
ASIC slot #3   -
ASIC slot #4   -
ASIC slot #5   46.0 ℃
ASIC slot #6   -

Did BertMod help you ? Please help back !
Donations are welcome to 1oHj13eEj1cctkUHzwWq38niBSX13WsSt
Thank you. -- uski

STATUS=S
When=1390229846
Code=11
Msg=Summary
Description=cgminer 3.9.0|SUMMARY
Elapsed=1436
MHS av=459462.58
MHS 5s=463149.23
Found Blocks=0
Getworks=105
Accepted=943
Rejected=0
Hardware Errors=521
Utility=39.40
Discarded=74
Stale=0
Get Failures=0
Local Work=164465
Remote Failures=0
Network Blocks=6
Total MH=659797170.9788
Work Utility=6419.33
Difficulty Accepted=150947.00000000
Difficulty Rejected=0.00000000
Difficulty Stale=0.00000000
Best Share=779416
Device Hardware%=0.3380
Device Rejected%=0.0000
Pool Rejected%=0.0000
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    0.822 V    32.3 A (26.6 W)
1    OFF    OK    11.5 V    0.823 V    31.6 A (26 W)
2    OFF    OK    11.5 V    0.82 V    32.2 A (26.4 W)
3    OFF    OK    11.5 V    0.823 V    32.4 A (26.7 W)
4    OFF    OK    11.5 V    0.823 V    33.3 A (27.4 W)
5    OFF    OK    11.5 V    0.821 V    33.3 A (27.3 W)
6    OFF    OK    11.4 V    0.823 V    32.2 A (26.5 W)
7    OFF    OK    11.4 V    0.819 V    32.8 A (26.9 W)
4    

Temperature sensor: 46.5 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    11.6 V    0.825 V    33 A (27.2 W)
1    OFF    OK    11.5 V    0.823 V    32.6 A (26.8 W)
2    OFF    OK    11.6 V    0.824 V    32.3 A (26.6 W)
3    OFF    OK    11.6 V    0.822 V    32.4 A (26.6 W)
4    OFF    OK    11.7 V    0.824 V    33.6 A (27.7 W)
5    OFF    OK    11.6 V    0.822 V    33.5 A (27.5 W)
6    OFF    OK    11.6 V    0.822 V    31.9 A (26.2 W)
7    OFF    OK    11.6 V    0.823 V    31.9 A (26.3 W)

Total DC/DC power output: 428 W


Next step would be immersion I think as I can no more avoid FAULT 4 if pushed more.

It has been running like this 2 days now. Added aluminium heatsinks to vrm, took the tape off before applying, cleaned everything up with acetone.  Arctic silver adhesive (expensive but its like glue), adjust the "income" fans so they are next to asics, leaning "down" pushing air through asic-boards and vrms. There wasn't any "heat transfer" rubber on the "front" vrm so cut half the back rubber thats on "last" vrm and inserted that in between the metal and vrm.

Tried reversing airflow on asics as I tought that it would be better to be sucked in rather than blown out. This didn't work.

Number of shares in this example is low because I once more tried to push it and had to restart.

I have driven 500gh/s for couple of minutes! Then shit starts to happen =). My worker wont show this because I split power to 2 pools, bitparking long round is annoying to say at the least.

Oil tank, I will be searching for you!

[temen]

Oh.. I think you would ask for it =)

It's November Saturn. Only 2 modules! Corsair RM 850  has started to use it's fan with this mod:)

I took off the metal behind the blowers in the front. Also the fans are repositioned.

gminer3.8.5 didn't do it this time. using 3.9.0.

Here it is. 305 is the code. Don't change the last digit. If you want more, start increasing 315 325 335 345 etc, until 3F5, after that 405 etc. I havent tried past 345.

If you want less use 2F5, then 2E5, then 2D5, 2C5 etc.. going down.

When chancing the value at cgminer.sh start always with /etc/init.d/cgminer.sh.

DON't START WHEN WARM! Let it cool at least 30 seconds.

If you blow it, its your fault. I have added heatsinks, modified blowers to different place etc.

donations accepted=)

BTC: 1BQq86NeKXY9j4qmVx4SKtGSvsEctgfove

Thanks in advance.

Someone knows how to change vrm output voltage from cgminer script?!  If I could lower the volts I could be running excess 500ghs!