Author

Topic: Swedish ASIC miner company kncminer.com - page 1161. (Read 3049528 times)

hero member
Activity: 742
Merit: 500
November 03, 2013, 08:22:38 AM
Actually I think Phoenix's method is more dangerous than mine because he is blocking the lower portion of the heatsink. I am blocking the top i.e. further away from the chip. The purpose of a heatsink to draw and dissipate heat away. My method allows the heat to be drawn away where it counts most and as long as the IR temp gun confirms 140F or below on the top of the chip it is probably quite safe.

Also noticed that. Although the back of the heatsinks are still completely free to radiate in his setup and the fans are blowing.

sr. member
Activity: 378
Merit: 250
November 03, 2013, 07:49:19 AM
Actually I think Phoenix's method is more dangerous than mine because he is blocking the lower portion of the heatsink. I am blocking the top i.e. further away from the chip. The purpose of a heatsink to draw and dissipate heat away. My method allows the heat to be drawn away where it counts most and as long as the IR temp gun confirms 140F or below on the top of the chip it is probably quite safe. System still running like a champ at 269 hashrate after days. It is not that odd to not have a fan on the heatsink because in the old days even the CPU did not have a heatsink. First they added heatsinks and then heatsinks with fans. The only goal is prevent overtemp of the chip. Oh, and I do work in engineering/IT. Designed software, hardware, tools etc. over the years. Just have not done much with Linux. I love it when people tell me "you can't do that" because I've proven them wrong more than once like when I wrote a program to default the password at the hex level in a program used in many police stations. The cops kept forgetting their password especially because it made you change it every so often.
hero member
Activity: 742
Merit: 500
November 03, 2013, 06:40:11 AM

just hope we don't see a "toboggan run" down this steep slope when everyone xfers their BTC to USD in a panic to get ROI back.
 

Why would anyone want to convert a promising sound currency for an exponentially-printed-to-oblivion/give-inflation-to-bankers-enslave-everyone-else/no-control-over-your-money corrupt currency?


Hmmm, so the temperature is measured on the outside of the plastic bags and cardboard that some are using to warm their chips?

Why do I smell burning ASICs?  Lol.

That is how the discussion started when I warned FiatKiller about it.  Smiley

But trusting Phoenix's values, the chips should hold a lot more than the VRMs.

copper member
Activity: 2898
Merit: 1465
Clueless!
November 03, 2013, 04:48:15 AM
Well according to the genesis block calculator http://mining.thegenesisblock.com/   difficulty has gone up 950% in the last 90 days. This does not bode well.....almost 500% the last 60 days...

just hope we don't see a "toboggan run" down this steep slope when everyone xfers their BTC to USD in a panic to get ROI back. Myself I'm so far in the hole with the BFL scam...I'm just gonna ride it out ...so at least I get a good story out of the whole thing.....'Well back when I was young and full of p*ss and vinegar and thought I was smart.....(you can guess how the rest goes) ....

fun ride...making good time....what could possibly go wrong right?

Searing
 
member
Activity: 82
Merit: 10
November 03, 2013, 12:57:46 AM
any jupiter owners running .98 notice that the hashrate is highest when you do a poweroff cycle and reboot, it will hit 565-575gh/s at the pool and then after running for a while the pool listed hash rate will get progressively lower, i dip at 490gh/s

any idea why this is?



This is exactly what happens to mine, maybe even lower, 470, 460... no idea why  Huh
did you ever try 70-75C? after enablecores, on 0.98?

24 hours later.. my 3 sats are 850 at the pool on 12 hour average!
that's 283 each, which is exactly what they show on cgminer!

what are your temp readings? how are your fans/case set up? im thinking of following you, doing enable core, and rebooting and then putting case on upside down to raise the temps a bit and see if i still get a down slide in hashes over time with higher temps.


at the moment, all 6 boards(3 sats) are between 69-73C, no case fans whatsoever, heatsink fans only. cases off, stacked, with cardboard partially blocking the fans airflow into the heatsink, like this...

All running 0.98 with enablecores over top
this is for night-time basically, because it gets downright chilly in here.
During the day, I can remove the cardboard, and they stay around 70
Ive tried several times to leave them off... but if they drop below those
temps by more than about 5 degrees, I see a drop in my rates, and can
see one one miner loose about 6 cores, and another loose two...
then they take forever to come back, like 6-8 hours!
I have one miner that is super-rock solid, one a tad finicky,
and one super-sensitive finicky, don't breathe near me kind
Also, It may just be my imagination, but they seem to work better
while its raining outside(high humidity)

thanks for sharing, phoenix!
i can confirm that after putting my case on entirely and raising the temp by about 8-9 degrees my hashrate has remain steady back in the 550-570gh/s on btcguild pool readings.
 
hero member
Activity: 784
Merit: 1004
Glow Stick Dance!
November 03, 2013, 12:52:33 AM
just confirming, yes.
But also good to know the temp sensor readings are from a different location on bertmod than the normal firmware...  perhaps the Bertmod Dev could add the asic temp, or KNC could add vrm temps....
 
Based on the vrm info.. I'm trying a few things....  details later.


Both normal firmware and bertmod are using data from lm75 temp sensors (chip in red circle in the picture below). One sensor is on each mining board and one is on controller board. Differences are probably caused by refresh rate of the temp data. Stock firmware refreshes data each minute, stores them to /var/run/stats.knc and displays this file. Bertmod refreshes when requested.

Unfortunately, there is no way how to get temperature from VRMs (or at least I did not found it in datasheet)




Hmmm, so the temperature is measured on the outside of the plastic bags and cardboard that some are using to warm their chips?

Why do I smell burning ASICs?  Lol.
hero member
Activity: 784
Merit: 1004
Glow Stick Dance!
November 03, 2013, 12:47:08 AM
Reporting back on FW 0.98.1

Jupiter 3 with 2 bad dies was hashing at 487 GH/s

20mins in with FW 0.98.1 I'm now hashing at 526 GH/s  Smiley

Still not 550- but I'll take it.


Thank you for your report. Anyone tried to upgrade already good miner?

I dont know if I have to risk messing it up

not me I'm the most confused and biggest coward on this thread...I re-started my miner 2 days ago and now it is running at 568gh and 540gh on slush 5.2 percent error rate and 7770 wu....on version .95..from my noob perspective from what I can tell that is pretty good...I've no idea why it went up like 13gh stable..temp in bsmt the same as is cores etc ..still  have top off and fans blowing crossways 5" fans that is both in front of unit...

absolutely opposite from what I can tell what phoenix is getting ...it is dry air in Minnesota however and cool again 62F bsmt and he is in Hawaii could humidity have something to do with all this?

anyway NOT  touching nothing....(I still avert eyes away from unit ..will keep evil jally bfl unit I got on Halloween (appropriate) at other end of the house...lest it jinx it...STILL think I should get voodoo for dummies and chicken entrails as good an explanation as any..in that the unit arrived in trashed box and full of rainwater)

so anyway maybe I could get insane GH upgrading to .98.....no balls to do so at this time

Searing


Smart man!  I have a feeling that some of these guys who can't stop experimenting with their gear are going to be very sorry when their miners have meltdown.  Wink
legendary
Activity: 938
Merit: 1000
LIR DEV
November 02, 2013, 10:24:02 PM
open, stacked

Q: Has anyone tried the bfgminer on eligius?
any difference?...stable?
able to set intensity?
ImI
legendary
Activity: 1946
Merit: 1019
November 02, 2013, 10:21:17 PM

Okay, I'm done fiddling..  This is 0.98 with enablecores again
heatsink fans on stock regular way, nothing else.



Funny thing is, the Saturn with the 3 disabled cores is the fastest, and has the lowest error rate...
I'm thinking these speeds are about as good as it gets without overclocking.

case open? case closed?
legendary
Activity: 938
Merit: 1000
LIR DEV
November 02, 2013, 10:18:53 PM

Okay, I'm done fiddling..  This is 0.98 with enablecores again
heatsink fans on stock regular way, nothing else.



Funny thing is, the Saturn with the 3 disabled cores is the fastest, and has the lowest error rate...
I'm thinking these speeds are about as good as it gets without overclocking.
hero member
Activity: 742
Merit: 500
November 02, 2013, 08:29:01 PM

Huh There's two temp sensors one is in the VRM itself, but some boards have an analogue DCDC which is why they may not show every statistic.

Uski brought this up the other day as well in this thread, and here post#65;

http://forum.kncminer.com/forum/main-category/main-forum/6183-bertmod-0-2-unofficial-firmware-mod-feedback-thread/page5

The reason you haven't had much of a comment from me is the only time time I have to check this thread currently is via cursory glances whilst on public transport, and even then the reception and a tiny mobile screen isn't favourable.

I wouldn't put solder to your boards if you value warranty btw.

In that post Uski is referring to the Amps, voltages and ON/OFF statuses that the digital DC/DC modules output. Not temperatures. Like you say they likely have a temperature sensor but that's likely only for the flags kill-switch mentioned.

Notice that in some of the screenshots people post of Bertmod, one can see the temperature sensor information but no DC/DC detected (certainly the analog ones like you say).  

I understand it might be hard to give anything other than the straight answers you believe in while keeping track in a phone. Sorry about complaining about deferment to authority. There is quite a bit of that here nonetheless and I feel I got dragged into this collective temperature sensor discussion/investigation. But do tell us about your conditions now. Will you be employed by KNC? Staying in Sweden for the November batch? Will you get a place to sit rather than your phone?

  
hero member
Activity: 532
Merit: 500
November 02, 2013, 08:13:37 PM

But also good to know the temp sensor readings are from a different location on bertmod than the normal firmware...  


They are not man...


Regarding temperature sensor data, I decided to do a simple test...

Dude I've said Bertmod gives VRM temps, as has Uski.

Can you clarify regarding NON Bertmod reported temperatures? I imagine they are the same but to not leave margin for doubt here. Thank you.



Non Bertmod is from a separate temp sensor.

ok, so what's the verdict here?  I'm just wondering ...  
Why you think O'rama is wrong?...
Your experiment results, or info from Bertmod DEV?

Simply because O'rama was just deferring to authority (putting things in the form of "I already said" and "It is"..) without comment to the conflicting experimental results.

When experimental results contradict deferment to authority, authority is usually flat out wrong, not just a little bit.

Thank you so much dzindra and kill-switch for not only finally settling this but for going through datasheets and what not to give us all this insight into temps!



Huh There's two temp sensors one is in the VRM itself, but some boards have an analogue DCDC which is why they may not show every statistic.

Uski brought this up the other day as well in this thread, and here post#65;

http://forum.kncminer.com/forum/main-category/main-forum/6183-bertmod-0-2-unofficial-firmware-mod-feedback-thread/page5

The reason you haven't had much of a comment from me is the only time time I have to check this thread currently is via cursory glances whilst on public transport, and even then the reception and a tiny mobile screen isn't favourable.

I wouldn't put solder to your boards if you value warranty btw.
member
Activity: 100
Merit: 10
November 02, 2013, 08:09:29 PM
according to the poster it works. Shocked
legendary
Activity: 1400
Merit: 1000
I owe my soul to the Bitcoin code...
November 02, 2013, 08:08:27 PM
Solder mod over here: https://bitcointalksearch.org/topic/reparacion-kncminer-saturno-fallo-dcdc-323718

Is it for real? Anyone tried it yet?

Looks like someone attempted a 'fix' for a bad DC/DC module.  Is it even possible to repair?
hero member
Activity: 742
Merit: 500
November 02, 2013, 08:06:00 PM
Solder mod over here: https://bitcointalksearch.org/topic/reparacion-kncminer-saturno-fallo-dcdc-323718

Is it for real? Anyone tried it yet?

Not a mod, a repair of a missing or bust DC/DC.

member
Activity: 100
Merit: 10
November 02, 2013, 08:00:49 PM
Solder mod over here: https://bitcointalksearch.org/topic/reparacion-kncminer-saturno-fallo-dcdc-323718

Is it for real? Anyone tried it yet?
newbie
Activity: 11
Merit: 0
November 02, 2013, 06:00:37 PM
Thank you so much dzindra and kill-switch for not only finally settling this but for going through datasheets and what not to give us all this insight into temps!

You are welcome Wink I have been digging in my miner since it arrived and finally I can put the gained knowledge to good use Smiley
hero member
Activity: 742
Merit: 500
November 02, 2013, 05:54:27 PM

But also good to know the temp sensor readings are from a different location on bertmod than the normal firmware...  


They are not man...


Regarding temperature sensor data, I decided to do a simple test...

Dude I've said Bertmod gives VRM temps, as has Uski.

Can you clarify regarding NON Bertmod reported temperatures? I imagine they are the same but to not leave margin for doubt here. Thank you.



Non Bertmod is from a separate temp sensor.

ok, so what's the verdict here?  I'm just wondering ...  
Why you think O'rama is wrong?...
Your experiment results, or info from Bertmod DEV?

Simply because O'rama was just deferring to authority (putting things in the form of "I already said" and "It is"..) without comment to the conflicting experimental results.

When experimental results contradict deferment to authority, authority is usually flat out wrong, not just a little bit.

Thank you so much dzindra and kill-switch for not only finally settling this but for going through datasheets and what not to give us all this insight into temps!

newbie
Activity: 11
Merit: 0
November 02, 2013, 05:30:38 PM
Both normal firmware and bertmod are using data from lm75 temp sensors (chip in red circle in the picture below). One sensor is on each mining board and one is on controller board. Differences are probably caused by refresh rate of the temp data. Stock firmware refreshes data each minute, stores them to /var/run/stats.knc and displays this file. Bertmod refreshes when requested.

Unfortunately, there is no way how to get temperature from VRMs (or at least I did not found it in datasheet)
https://i.imgur.com/HFZZ6t9.jpg


Awesome!  There is temp data available if the VRM is the 'digital' version, the details are on page 16 of the datasheet.  After a close read, it does seem you are indeed correct, that the PMBus will only report two temperature 'flags' OT_WARN and OT_FAULT, not actual temperature data.  The location of the lm75 makes sense why it seems to track the DC/DC module (PCB) temperatures better than reporting accurately for the ASIC itself.

Exactly. And it also seems that these limits are hardwired into VRM (so no way of lowering them and checking warn flags to get at least approximate temp value).
newbie
Activity: 11
Merit: 0
November 02, 2013, 05:11:57 PM
just confirming, yes.
But also good to know the temp sensor readings are from a different location on bertmod than the normal firmware...  perhaps the Bertmod Dev could add the asic temp, or KNC could add vrm temps....
 
Based on the vrm info.. I'm trying a few things....  details later.


Both normal firmware and bertmod are using data from lm75 temp sensors (chip in red circle in the picture below). One sensor is on each mining board and one is on controller board. Differences are probably caused by refresh rate of the temp data. Stock firmware refreshes data each minute, stores them to /var/run/stats.knc and displays this file. Bertmod refreshes when requested.

Unfortunately, there is no way how to get temperature from VRMs (or at least I did not found it in datasheet)

IMAGE EDITED OUT
thanks...
guess FeedBackLoop may be right!, right?

Yes, temperature readings are measured by the same sensors but timing is critical here. You can have slightly different readings each time you measure temperature.

If you are interested and want a proof, ssh into your box, open /sbin/monitordcdc, read through it and look for following line (present at line 42 in 0.98):

Code:
temp1=$(i2cget -y $bus 0x48 0 w 2>/dev/null)

This line reads two bytes from miner board temp sensor. You can find same command being issued in asic_status.pl (bertmod0.3, line 37)
Jump to: