Hacking The KNC Firmware: Overclocking

[temen]

Your boards are november, mine are october, i don't think they can be pushed as far.

I was more worried about the PCI cables and extenders, but they are nice and cool.

quote

For me i couldn't change almost any values at all before I added heat sinks on the vrms, repositioned the fans etc. Those mods made huge difference. It owould be best to get airflow underneath the vrms also as the sink heat to pcb. Someone said that heatsinks dont add so much but for me it was turning point.

Of course octobers are different and one would want to put full power immediately on, because this way the boards are under great stress. Make the boards warm first before applying "full" power.

For october batch I would consider oil-immersion and injecting voltage after vrms to board. Like getting some huge dc-dc transforming device and using it to power asics. Same time boards sunken into liquid-cooling system.

Lets hope that everybody can max out their hahsrates! But remember: there is a reason why boards are not overclocked from the factory. If you change something and it blows its your fault. dont ignite any fires etc. Safety first, mining third!

[1714038355]

1714038355

[1714038355]

1714038355

[1714038355]

1714038355

[1714038355]

1714038355

[elasticband]

I have been running 231 for a few weeks, then 271 for the last few days. I guess i will let you all know how i get on tomorrow morning.

I do have extra cooling and temps range from 41 - 48, i do think over 200 is pushing it. I am trying to tune it down slightly but it comes at a sacrafice of hashrate

still running 291, 810w 790GH, that is a 240GH increase from stock, almost 45%increase from stock

[DPoS]

I have been running 231 for a few weeks, then 271 for the last few days. I guess i will let you all know how i get on tomorrow morning.

I do have extra cooling and temps range from 41 - 48, i do think over 200 is pushing it. I am trying to tune it down slightly but it comes at a sacrafice of hashrate

still running 291, 810w 790GH, that is a 240GH increase from stock, almost 45%increase from stock

nice..  I got some chip heatsinks this week, so I'll try to up it to at least 251 today

[bondus]

A warning:

For you brave souls that have enabled all 8 VRMS (early Octobers) I saw today that the 1.0.0 firmware does not apply the voltage adjustment to all 8 VRMs. It only applies it to the 4 that are officially enabled. The old FW applied this adjustment to all VRMs.

The VRMs work in pairs and if the voltage adjustment (VOUT_TRIM register) of the 2 VRMs in a pair is different things will go bad.

If you still want to use all 8 VRMs I suggest you read and understand the datasheet of the VRM.
The VRMs are very resilient to abuse, but they have limits.


Other than that FW 1.0.0 works very good. HW errors have dropped a lot for me. Good chips are close to 0% errors (2% before), bad chips still close to 2%.

[sickpig]

A warning:

For you brave souls that have enabled all 8 VRMS (early Octobers) I saw today that the 1.0.0 firmware does not apply the voltage adjustment to all 8 VRMs. It only applies it to the 4 that are officially enabled. The old FW applied this adjustment to all VRMs.

The VRMs work in pairs and if the voltage adjustment (VOUT_TRIM register) of the 2 VRMs in a pair is different things will go bad.

If you still want to use all 8 VRMs I suggest you read and understand the datasheet of the VRM.
The VRMs are very resilient to abuse, but they have limits.


Other than that FW 1.0.0 works very good. HW errors have dropped a lot for me. Good chips are close to 0% errors (2% before), bad chips still close to 2%.

many thanks bondus. very helpful as usual.

So how did you manage to apply the advanced settings to all the VRMs after upgrading to the new firmware? did you just use the old fetch_advanced_settings.cgi or did you also use the old waas executable?

[jelin1984]

Any details how we can over clocked October Jupiter with new 1.00 firmware ?

[ncs0ne]

Any details how we can over clocked October Jupiter with new 1.00 firmware ?

I'm running the same script as I did with the previous firmware.
Works fine as usual.

[j03]

This 1.0 firmware - KnC states you can change clock speeds through the GUI. Can you actually overclock on a per die basis? They were always talking against overclocking and voiding the warranty. Can someone using 1.0 confirm this? Screenshot? Thanks.

[jelin1984]

Can you post again the details and the code please

Also I can delete and copy the code?

[kendog77]

This 1.0 firmware - KnC states you can change clock speeds through the GUI. Can you actually overclock on a per die basis? They were always talking against overclocking and voiding the warranty. Can someone using 1.0 confirm this? Screenshot? Thanks.

Very interesting. I didn't notice this before, but it looks like the default clock of an October Jupiter is 750, and it can be increased to 775 in the 1.0 web GUI.

The default clock of a November Jupiter is 900, and it can be increased to 1000 in the 1.0 web GUI.

I guess it's time to play around now...  

[jelin1984]

How we can do that?

Which option button..

[kendog77]

How we can do that?

Which option button..

Under the "Advanced" tab, each ASIC die now has a pull down menu that allows the user to adjust the chip frequency. It's right beside the voltage pull down button.

[j03]

This 1.0 firmware - KnC states you can change clock speeds through the GUI. Can you actually overclock on a per die basis? They were always talking against overclocking and voiding the warranty. Can someone using 1.0 confirm this? Screenshot? Thanks.

Very interesting. I didn't notice this before, but it looks like the default clock of an October Jupiter is 750, and it can be increased to 775 in the 1.0 web GUI.

The default clock of a November Jupiter is 900, and it can be increased to 1000 in the 1.0 web GUI.

I guess it's time to play around now...  

Ok only to 775 via the GUI on a October Jupiter. That makes sense. They're not letting people do much. Custom hacks are still the way to go then. Thanks.

[crashoveride54902]

A warning:

For you brave souls that have enabled all 8 VRMS (early Octobers) I saw today that the 1.0.0 firmware does not apply the voltage adjustment to all 8 VRMs. It only applies it to the 4 that are officially enabled. The old FW applied this adjustment to all VRMs.

The VRMs work in pairs and if the voltage adjustment (VOUT_TRIM register) of the 2 VRMs in a pair is different things will go bad.

If you still want to use all 8 VRMs I suggest you read and understand the datasheet of the VRM.
The VRMs are very resilient to abuse, but they have limits.


Other than that FW 1.0.0 works very good. HW errors have dropped a lot for me. Good chips are close to 0% errors (2% before), bad chips still close to 2%.

THanks to his help via IRC I was able to update fw and still use my 2 8vrm boards...and my hw error droped down to .11% from .25% yay hopefully it stays that low as i just restarted...Thxs bondus

[arousedrhino]

Your issues sound somewhat similar to mine. I did the SD card reset. It works, but the machine seems to like me messing around with it less and less, because the "can't browse to machine IP" issue happened again relatively quickly.

How to do a SD-Card reset? Seems I have a bricked KnCMiner...

Hello,

I seemed to have that problem in early January. Following the guide here http://forum.kncminer.com/forum/main-category/hardware/23664-how-to-restore-a-broken-beaglebone-without-the-need-to-rma really helped me out.

If you have a mac you don't need to get an Ubuntu Live USB/CD, if you don't have a mac or linux PC then you need to get the Ubuntu Live USB/CD to format the micro-sd card.

Also make sure you micro-sd card is under 4 GB (Tip: Radioshack does not have them, Walgreens does though) 

Best of luck getting it back up and running!

[elasticband]

still running 291/790GH/810w oct. jupiter 

[sickpig]

A warning:

For you brave souls that have enabled all 8 VRMS (early Octobers) I saw today that the 1.0.0 firmware does not apply the voltage adjustment to all 8 VRMs. It only applies it to the 4 that are officially enabled. The old FW applied this adjustment to all VRMs.

The VRMs work in pairs and if the voltage adjustment (VOUT_TRIM register) of the 2 VRMs in a pair is different things will go bad.

If you still want to use all 8 VRMs I suggest you read and understand the datasheet of the VRM.
The VRMs are very resilient to abuse, but they have limits.


Other than that FW 1.0.0 works very good. HW errors have dropped a lot for me. Good chips are close to 0% errors (2% before), bad chips still close to 2%.

THanks to his help via IRC I was able to update fw and still use my 2 8vrm boards...and my hw error droped down to .11% from .25% yay hopefully it stays that low as i just restarted...Thxs bondus

do you care to share bondus help?

[jelin1984]

how i can delete the code
and put the new one code?

any new code

is safe to have amper at 50 a or will detsroy it?


Miner Status
Mining Status

CGMiner Status   Running (pid=19609)
Last Checked   Fri Feb 21 12:44:19 UTC 2014
Avg. Hash Rate   645 Gh/s
Difficulty Accepted   44,096
Difficulty Rejected   352 (0.8 %)
Hardware Errors   704 (1.5 %)
WU   9,025
Difficulty Stale   0
Network Blocks   3
Pool Rejected   0.8 %
Pool Stale   0
Best Share   7,434
Found Blocks   0
HW Status

ASIC slot #1   52.0 ℃
ASIC slot #2   -
ASIC slot #3   56.5 ℃
ASIC slot #4   55.0 ℃
ASIC slot #5   40.5 ℃
ASIC slot #6   -
ASIC Board Status
Update ASIC Status
Last updated on Fri Feb 21 12:17:41 2014

ASIC Board   Info
0   
Temperature sensor: 33.0 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   ON   OK   11.8 V   0.782 V   40.1 A (31.4 W)
1   No DC/DC detected
2   ON   OK   11.8 V   0.798 V   39.8 A (31.8 W)
3   No DC/DC detected
4   ON   OK   11.9 V   0.785 V   41.2 A (32.3 W)
5   No DC/DC detected
6   No DC/DC detected
7   ON   OK   11.8 V   0.771 V   41.2 A (31.8 W)
2   
Temperature sensor: 35.5 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   ON   OK   11.8 V   0.751 V   41.1 A (30.9 W)
1   No DC/DC detected
2   ON   OK   11.9 V   0.781 V   42 A (32.8 W)
3   No DC/DC detected
4   ON   OK   12 V   0.789 V   41.6 A (32.8 W)
5   No DC/DC detected
6   No DC/DC detected
7   ON   OK   11.8 V   0.804 V   42.4 A (34.1 W)
3   
Temperature sensor: 35.5 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   ON   OK   11.9 V   0.783 V   40.7 A (31.9 W)
1   No DC/DC detected
2   ON   OK   11.8 V   0.748 V   41.9 A (31.3 W)
3   No DC/DC detected
4   ON   OK   11.9 V   0.778 V   42.6 A (33.1 W)
5   No DC/DC detected
6   No DC/DC detected
7   ON   OK   12 V   0.785 V   41.5 A (32.6 W)
4   
Temperature sensor: 31.0 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   ON   OK   11.9 V   0.789 V   40.4 A (31.9 W)
1   No DC/DC detected
2   ON   OK   11.8 V   0.792 V   40.8 A (32.3 W)
3   No DC/DC detected
4   ON   OK   11.9 V   0.783 V   41.4 A (32.4 W)
5   No DC/DC detected
6   No DC/DC detected
7   ON   OK   11.9 V   0.808 V   42.2 A (34.1 W)
Total DC/DC power output: 517 W

[Powell]

Got my second October 4VRM Jupiter in and on first bootup worked fine for a few hours.  I updated to 1.0 to test and I noticed when I turned off SPI Auto this:



I've tried the hit it with SPI 3.3 and max voltage.  Let it run a few hours like that.  I have had them drop when OC to high but this I can't seem to rescue.



There is it at max voltage on that one.  I've swapped back to .99.1 , .99.2 and 1.0 and can't seem to get it to jump back up.

I've tried running 231 and 221 but sadly nothing helped kick that one up even. 

Any other suggestions so I can put this Jupiter to a happy #

[j03]

still running 291/790GH/810w oct. jupiter 

291?! Hot damn!