This is mine.

I spent long hours when it arrived to figure out wich pci-e slot works fine with which hash card.

I've replaced the stock fans with 2 110cfm, 2 150cfm and at the intake in the middle there is one 150cfm. 2 170 cfm intake fans are on its way.

Now the result is this.

Room temp around 19 c.

My hashrate on Btcguild between 520-570Ghash

Now I'm trying to keep my hands off the box. If it ain't broke...

not comfortable applying 0.96

but willing to try edit/config/cgminer.conf

where is this? in the KnC gui or via PuTTy?

OUCH! BTCGuild is down fluctuating, how do we apply a failover pool to these rigs?

I just did some port forwarding on my router.  Now i can access my KNC page from my cell phone when I'm out to check on it,update or reset, Also set up JuiceSSH app on my android.  So now i can connect to cgminer through ssh on the phone as well.  
https://i.imgur.com/GmfeKXD.png
https://i.imgur.com/QyvVugC.jpg?1

Yeah, great idea!
(Don't do it in an earthquake region!)

Here is my setup. I removed that external case for the fans.
The results are great. The BertMod (modified ASCII script) reported temperatures dropped by 4-5C:

Old setup (with two additional fans on the side, one on top):
Board 0: Temperature sensor: 40.5C
Board 3: Temperature sensor: 40.5C

Modified Board 3 (the right one looking from the BBB):
Board 0: Temperature sensor: 41.0C
Board 3: Temperature sensor: 37.5C

Modified Board 0 as well and waited 10 minutes:
Board 0: Temperature sensor: 36.5C
Board 3: Temperature sensor: 36.0C

This setup also cools the VRMs much better. I had two ones (same location, see picture!) with a temperature of
51-52C. Those are now at about 47C. The only side effect is, that the heatsink is getting warmer. But - it's designed
for that. See the infrared thermometer measurements on the picture.

http://s23.postimg.org/fi2kq5plj/Saturn_Cooling_v4a.jpg
http://s17.postimg.org/mpj2had3f/Saturn_Cooling_v4.jpg
http://s17.postimg.org/p9epbe0nf/Saturn_Cooling_c.jpg

No temperature reading for the 4th VRM, sorry!

trepex

which is why I lowered the fans to get 5-7C cooler  

Good work!!  

 and yes another way to keep cgminer from starting and turning off cores is to just change the first { to a } in the cgminer.conf so it runs but just sits there (screen -r and just quit it)

my one die being off is probably saving enough watts so my 850psu doesnt roast...  I'll pick up a 1000W or 1200w tomorrow


Edit:

3rd time was the charm!

I'm not sure I'm running the same version of cgminer, 3.4.0 from the top line.  My cgminer doesn't note the temp.

Interesting... applied 9.6.1 and it still shows 9.6 and initially everything was worse than 9.6

So I applied enablecores again and too early to say for sure, but looking very good.

I've also noticed an interesting phenomena of "bad boots". Sometimes after a restart things are not up to par, but if you keep trying you will hit a "golden boot" that works much better. I recommend for people having trouble to try this!

Never seen this much speed before, but too early to say how it will stabilize:

https://dl.dropboxusercontent.com/u/40920348/KNC961EC.jpg

Firmware 0.9.6.1 just released

The die that came back was Die "1" on my slot 4, below is a current bertmod output of it. I have 12 other cores out across the other chips for 13 total now and usually they almost all turn on after 2 hours. I still have this damn power down slide after 6 hours but it slides only to 500 ghs in the five second instead of 440ghs or less as it use to before enablecores.bin usage.  

(5s):541.7G (avg):519.5Gh/s NOW!

Temperature sensor: 38.5 C

Die ID   Cores ON   Cores OFF   %
0   48   0   100
1   47   1   97.9
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   0.744 V   35.5 A (26.4 W)
1   OFF   OFF   11.8 V   0.712 V   0 A
2   ON   OK   11.8 V   0.738 V   38.5 A (28.4 W)
3   OFF   OFF   11.8 V   0.747 V   0 A
4   ON   OK   12 V   0.745 V   38.2 A (28.5 W)
5   OFF   OFF   11.9 V   0.695 V   0 A
6   OFF   OFF   11.8 V   0.739 V   0 A
7   ON   OK   11.8 V   0.732 V   40.3 A (29.5 W)
Total DC/DC power output: 446 W

#!/bin/sh
for i in 0 1 2 3 4 5 ; do
        ./enable_all_cores $i >/dev/null 2>&1
        if [ $? = 0 ] ; then
                echo ASIC board \#$i: All cores enabled\</br\>
        else
                echo ASIC board \#$i: not found\</br\>
        fi
done
sync

Hi,

I just noticed something very interesting while trying to get bfgminer work:

I "q"uit cgminer and the screen session terminated.
I then started cgminer manually with

# cgminer -c /config/cgminer.conf

As a result the BertMod tool is not working anymore but on the other hand, the performance did go up
like 4GH (on a Saturn) to 274.4-275. A value that I never managed to see stable for a long time.
Of course I don't have any long-term experience yet. Maybe someone else wants to try this.

cgminer is currently not running in a screen session but directly. Might be a reason for the better performance as well.

That would put the reported DC output of the module higher (430W ) than the computing DC input (385W).  Something isn't correct.   So either your wall wattage numbers or the output reported by the VRM is incorrect, they both can't be right.  Under ideal conditions (no cooling or host power consumption), 90% DC efficiency, 93% ATX PSU efficiency.  430W output would mean (430/(0.9*0.93) = 513W) >500W at the wall.

Power Supply model.

Thanks for the datapoints.  It is strange the reported DC/DC output doesn't change.

~430W out regardless however the input wattage changes significantly.

Need to make some assumptions but lets say the 6 fans use 6W ea (someone can look at the fan sticker) and the host uses another 5W.   So balance of system is ~40W @ 12VDC.  Lets also assume your PSU is 90% efficient at

v0.90 = 1650W @ 220VAC ~= 1485W DC @ 12VDC  (1485 - 40)/2 =  722W.
v0.95 =  946W @ 220VAC ~= 850W DC @ 12VDC  (850-40)/2 = 405W.

v0.90 VRM In: 722W Out: 433W Efficiency: 60% OUCH
v0.95 VRM In: 405W Out: 438W Efficiency: IMPOSSIBLE.  

So either your numbers or incorect or the output reported by the VRM is incorrect.   It is possible the PSU efficiency in the second case was slighly higher (say 92%) and the host power usage is less but those numbers don't change things significantly.  Looking at it the other way the VRM is at most 90% efficient.  438W out = 486W in (@ 90 efficiency) or 962W @ 12VDC for both.  Even with no fans or host wattage the AC wattage doesn't match the reported VRM output wattage even under ideal conditions (90% efficiency @ VRM, 92% efficiency @ PSU).    

Update (improved) KNC to 1.0 J/GH based on reported 455W @ 12 VDC running at 502 GH/s.  Assume 90% AC power supply efficiency results in estimated 505W at the wall.  Also moved KNC into the "actual devices" category.  Bitfury still holds the efficiency crown though.

https://bitcointalk.org/index.php?topic=170332.msg3307091#msg3307091

On edit: revised slightly to 1.1 J/GH.  The 455W was the output (0.75V) of the DC to DC converter.  Based on GE spec sheet the input current would be ~10% higher.  That puts it closer to 560W at the wall.  This appears to be inline with other customer reports using kill-a-watt type meters.  On some firmware and with the 4 VRM model the wattage is significantly higher ~1.3 J/GH.

 

your theory seems to be true though..  my one board with the dead die runs significantly hotter than the rest and that is with it hashing at 75%

can't see anything with the eyes that stands out, and while it hashes at 75% not really worth it to keep shutting things down to fudge with it for now

maybe RMA it when KNC gets out of the woods

Correct. I had e-mailed them regarding the status of my order(s) and had inquired on the performance issues people were having and Emilia stated a new FW will be on their site later today.

According to KnC, yep

Yup! I've my PSU unboxed, cables ran, firmwares downloaded (0.97 being released today  ), list of hardware checks, immediate fan mod- I'm ready to roll.