Author

Topic: Swedish ASIC miner company kncminer.com - page 1210. (Read 3049524 times)

legendary
Activity: 1066
Merit: 1098
October 26, 2013, 01:31:01 PM
Payout que is like this...
set the threshold....
balance reaches threshold, enters que
que is 8-10 blocks long
end of que, you get paid everything you've earned to that point.
simple.

Wow, are you really getting queued 8-10 blocks back???

I don't think I have ever had more than 1 or two blocks delay on there.

I know there are plenty of circumstances where a fail-safe will kick in and stop payouts completely until Wizkid restarts it manually, and payout will get backed up then - but when he restarts it, he manually does the payout also, so you don't have to wait for the queue to clear.

it seems larger quantities take a bit longer..... but yes..it varies..  my payments over a btc take a day, while small ones are within 4 -5 blocks... I said 8 to save any frustration... Smiley

If there is more than 25BTC worth of payout due, the payouts are done in order of how old the balance is.  If your payout is set too low, where you get paid very frequently, you will get queued when there is more than one block's worth of BTC to be paid out.  If you set it to get 1 or two payouts a day, you should not find yourself 8-10 blocks back.

legendary
Activity: 938
Merit: 1000
LIR DEV
October 26, 2013, 01:20:45 PM
Payout que is like this...
set the threshold....
balance reaches threshold, enters que
que is 8-10 blocks long
end of que, you get paid everything you've earned to that point.
simple.

Wow, are you really getting queued 8-10 blocks back???

I don't think I have ever had more than 1 or two blocks delay on there.

I know there are plenty of circumstances where a fail-safe will kick in and stop payouts completely until Wizkid restarts it manually, and payout will get backed up then - but when he restarts it, he manually does the payout also, so you don't have to wait for the queue to clear.

it seems larger quantities take a bit longer..... but yes..it varies..  my payments over a btc take a day, while small ones are within 4 -5 blocks... I said 8 to save any frustration... Smiley
legendary
Activity: 938
Merit: 1000
LIR DEV
October 26, 2013, 01:18:25 PM
wtf is there work week tuesday,wednesday, thursday. and thats it?  Angry

We made them all rich, what would you expect them to work every day now?


I imagine the new information on temps has them testing everything....
I also think they may be done manufacturing, and tying up loose ends....
datacenter mostly probably...
Party to follow!
legendary
Activity: 1066
Merit: 1098
October 26, 2013, 01:16:15 PM
Payout que is like this...
set the threshold....
balance reaches threshold, enters que
que is 8-10 blocks long
end of que, you get paid everything you've earned to that point.
simple.

Wow, are you really getting queued 8-10 blocks back???

I don't think I have ever had more than 1 or two blocks delay on there.

I know there are plenty of circumstances where a fail-safe will kick in and stop payouts completely until Wizkid restarts it manually, and payout will get backed up then - but when he restarts it, he manually does the payout also, so you don't have to wait for the queue to clear.
legendary
Activity: 1274
Merit: 1000
Personal text my ass....
October 26, 2013, 01:15:44 PM
wtf is there work week tuesday,wednesday, thursday. and thats it?  Angry

We made them all rich, what would you expect them to work every day now?

legendary
Activity: 938
Merit: 1000
LIR DEV
October 26, 2013, 01:12:44 PM
Payout que is like this...
set the threshold....
balance reaches threshold, enters que
que is 8-10 blocks long
end of que, you get paid everything you've earned to that point.
when threshold is met, you enter que again...
simple.
legendary
Activity: 1066
Merit: 1098
October 26, 2013, 01:12:23 PM


I went back to the Guild where I can at least input the amount of BTC I want to auto pay out



You can do that on Eligius too.  It's settable from your 'My Eligius' page.
legendary
Activity: 1274
Merit: 1000
Personal text my ass....
October 26, 2013, 01:08:38 PM
Yea, the graphs and things are cool like I said, but having no idea when I would be getting paid or what my 24 hour return is (if I'm making money or not) is a weird way to hash. At least with BTC they show you right on your stats page what you are making within a 24hour period, payout info, all the necessary info there is. Eligius is more of a guess right now, no? Also, if I wanted a manual payout with Eligius I can't because they don't have it. Bitcoin raised up to $201 yesterday and I wanted to sell a few bitcoin so I did a manual payout with BTCGuild, could I have done that with Eligius or would I have to wait 3 weeks until their payout schedule catches up or whatever? Yea, confusing at the least. I guess that's where the BTC fee comes in. You are furnished with all the appropriate information almost on one page.

you just 25*(Hashrate/2^32)/difficulty  (that's per second)

You get 25 bitcoins in a block
in order to find a block, you need to hash (on average) n diff1 shares, where n is the difficulty
in order to find a diff1 share, you need to calculate (on average) 2^32  hashes.  

2^32 = the number of bytes in a 4 gigabytes.

Alternatively, you can just go look at the PPS rate on btcguild, and assume you'll be getting a few % more then that.

That's the thing here....we are using computers that can do millions of calculations a second, I don't really want or think it is necessary for us to do math anymore when we have the best invention in the world that does it perfectly. Yea, I know...some of you still prefer to do your own calculations and expect everyone else to....but if I have a computer I kind of expect it to do all the math it can for me. I know, I know, that's why we are all dumb and stupid now. I would have just figured a mining pool would have all the important/critical info right available to the user instantly. I still couldn't figure out Eligius's payout schedule/process (yea, I guess I'm an ididot) so I went back to the Guild where I can at least input the amount of BTC I want to auto pay out. I know wizkid is doing his best, but when I see literally wizkid's name in the URL itself I have to chuckle a little. The main reason why I wanted to switch to Eligius anyway was I got very good hash rates there and no fees.

legendary
Activity: 938
Merit: 1000
LIR DEV
October 26, 2013, 12:57:45 PM
283 is awesome...don't touch a thing!   BAM
again, i saw the asic 28nm article hypermega posted....   pretty much explains it.
http://siliconsaint.blogspot.com/2012/07/temperature-inversion-in-deep-sub.html?m=1
legendary
Activity: 1036
Merit: 1000
Nighty Night Don't Let The Trolls Bite Nom Nom Nom
October 26, 2013, 12:49:44 PM
i like this bit  Cheesy:

Quote
KnCMiner products are carefully engineered and are not designed to be user serviceable in hardware or software other than when specifically directed by a member of the KncMiner team.
You will invalidate your product warranty if you try to disassemble any of the hardware or software, tamper with the running software in memory or otherwise, attempt to under or over-clock the device or interfere in anyway that is in contradiction to the intended use of the product.
Any performance altering changes to the hardware or software are logged/stored and will be used to validate any warranty claims
legendary
Activity: 1066
Merit: 1098
October 26, 2013, 12:49:27 PM
this "higher temp =  higher performance" doesnt make a lot of sense. Design aside, there are only two factors affecting performance:
- clockspeed. Is this variable on KnC? Can it be measured?
- number of functional cores/ % of HW errors. Does this go down as temps go up?

Thats it. The rest is statistical noise.

sense or not. i have one jupiter that reacts in the same way. additional cooling with a big box-fan on top of the miner = cgminer crashes every 5 min. without box-fan = cgminer stable / higher avg GH.

Can it be electrical interference from the fan's motor? They have quite a strong electromagnetic field when running.

It's probably a effect called temperature inversion. 28nm logic gets faster with higher temperatures.
http://siliconsaint.blogspot.com/2012/07/temperature-inversion-in-deep-sub.html?m=1

I don't think so...  That phenomenon might make it possible to clock a chip faster at a higher temperature, but it should not effect the speed of a chip that is clocked at a constant speed.  Right?



I honestly think it's a VRM thing... not an asic thing..  Smiley

I would think that as long as the voltage from the VRM is above the chip's lower operating threshold, it should not make any difference, unless they are being overclocked heavily.
could it be that the VRMs operate more efficiently at slightly higher temps and automatically provide slighlty higher amps without software/firmware instruction to do so....I do notice that the under performing chip has 2 cores with lower output volt. and current.....and that actually changes just by running the a/c and/or turning it off

p.s. saturn is on a apc ups, so can't be wall flucs

The purpose of a voltage regulator is to provide a constant voltage at any current drawn by the device it is powering, within it's rated limits.  The actual amount of current that flows is determined by the device it is driving, not the regulator.  Changing the amount of current the regulator is capable of providing when it is not near its limits would have no effect on operation.


newbie
Activity: 59
Merit: 0
October 26, 2013, 12:45:12 PM
I Also found that keeping them warm has made a huge improvement for me.   Last few nights i would wake up in the morning to find nice low temps in the mid 30c's but hasrate went down to around 538 or so  and hardware errors would  be over 11%  Then the hash rate on btcguild would end up in the low 400s ever morning.   So every morning i would start to re flash back to 95 and messing with enable cores and what not.  Keep messing with it.  And always seem around noon or so i would get it running better.  usually around  548 with around 9.9 error rate, and btcguild would show around 490 -505  So yesterday i deiced  to shut the door of the room the miner was in not letting the cool air come in from the night.    And this morning temps where around 44-52c depending on the board.  Hash rate is at 552 and erros down to 7.7% BTC quild showed 534 this morning.   And only two cores are off on one asic and all the other are 100% on   and before this i would have a total of between 6-8 cores off.  

So i don't care what anyone says.  I know for a fact mine runs allot better at a warmer temperature.
legendary
Activity: 1066
Merit: 1098
October 26, 2013, 12:42:09 PM
this "higher temp =  higher performance" doesnt make a lot of sense. Design aside, there are only two factors affecting performance:
- clockspeed. Is this variable on KnC? Can it be measured?
- number of functional cores/ % of HW errors. Does this go down as temps go up?

Thats it. The rest is statistical noise.

sense or not. i have one jupiter that reacts in the same way. additional cooling with a big box-fan on top of the miner = cgminer crashes every 5 min. without box-fan = cgminer stable / higher avg GH.

Can it be electrical interference from the fan's motor? They have quite a strong electromagnetic field when running.

It's probably a effect called temperature inversion. 28nm logic gets faster with higher temperatures.
http://siliconsaint.blogspot.com/2012/07/temperature-inversion-in-deep-sub.html?m=1

I don't think so...  That phenomenon might make it possible to clock a chip faster at a higher temperature, but it should not effect the speed of a chip that is clocked at a constant speed.  Right?




Not if the actual speed is already at the edge for the selected core supply voltage. In this case hardware errors can be caused by timing violations. Higher temperature, less timing violations, less hardware errors/crashs.

That's a valid point, but these chips are running at 45-50C for the most part.  They don't seem to be overclocked or stressed at all.

I guess I got lucky - my Saturn runs on all cores all the time - and I can vary the temperature over a pretty broad range by blowing hot air into the case with an old hair dryer, and I don't see any significant variation in hash rate.  Rock solid at 284.5 GH/s for the last 12 hours or more.  Haven't really tried any specific extra cooling.

That being said, I don't think it's impossible that whatever defect it is that makes the bad chips bad could be something that reacts to temperature.  I just don't think the phenomena is explained by temperature inversion.

newbie
Activity: 15
Merit: 0
October 26, 2013, 12:32:56 PM
Good point, thanks.

I asked Liam over the phone and he told me max temp for the ASIC chips is 105C, so you are far from the point where they are overheating Wink
legendary
Activity: 938
Merit: 1000
LIR DEV
October 26, 2013, 12:27:26 PM
Yeah, I noticed all this when looking at my graph on eligius, and correlating speeds with time of day & temps. It only took a few minutes to tell the fan was causing problems... the rest is history
In fact, the machines may be overcooled.
The results are undeniable.
sr. member
Activity: 336
Merit: 250
October 26, 2013, 12:23:03 PM
this "higher temp =  higher performance" doesnt make a lot of sense. Design aside, there are only two factors affecting performance:
- clockspeed. Is this variable on KnC? Can it be measured?
- number of functional cores/ % of HW errors. Does this go down as temps go up?

Thats it. The rest is statistical noise.

sense or not. i have one jupiter that reacts in the same way. additional cooling with a big box-fan on top of the miner = cgminer crashes every 5 min. without box-fan = cgminer stable / higher avg GH.

Can it be electrical interference from the fan's motor? They have quite a strong electromagnetic field when running.

It's probably a effect called temperature inversion. 28nm logic gets faster with higher temperatures.
http://siliconsaint.blogspot.com/2012/07/temperature-inversion-in-deep-sub.html?m=1

I don't think so...  That phenomenon might make it possible to clock a chip faster at a higher temperature, but it should not effect the speed of a chip that is clocked at a constant speed.  Right?



I honestly think it's a VRM thing... not an asic thing..  Smiley

I would think that as long as the voltage from the VRM is above the chip's lower operating threshold, it should not make any difference, unless they are being overclocked heavily.
could it be that the VRMs operate more efficiently at slightly higher temps and automatically provide slighlty higher amps without software/firmware instruction to do so....I do notice that the under performing chip has 2 cores with lower output volt. and current.....and that actually changes just by running the a/c and/or turning it off

p.s. saturn is on a apc ups, so can't be wall flucs
full member
Activity: 129
Merit: 100
October 26, 2013, 12:22:52 PM
this "higher temp =  higher performance" doesnt make a lot of sense. Design aside, there are only two factors affecting performance:
- clockspeed. Is this variable on KnC? Can it be measured?
- number of functional cores/ % of HW errors. Does this go down as temps go up?

Thats it. The rest is statistical noise.

sense or not. i have one jupiter that reacts in the same way. additional cooling with a big box-fan on top of the miner = cgminer crashes every 5 min. without box-fan = cgminer stable / higher avg GH.

Can it be electrical interference from the fan's motor? They have quite a strong electromagnetic field when running.

It's probably a effect called temperature inversion. 28nm logic gets faster with higher temperatures.
http://siliconsaint.blogspot.com/2012/07/temperature-inversion-in-deep-sub.html?m=1

I don't think so...  That phenomenon might make it possible to clock a chip faster at a higher temperature, but it should not effect the speed of a chip that is clocked at a constant speed.  Right?




Not if the actual speed is already at the edge for the selected core supply voltage. In this case hardware errors can be caused by timing violations. Higher temperature, less timing violations, less hardware errors/crashs.
legendary
Activity: 938
Merit: 1000
LIR DEV
October 26, 2013, 12:22:25 PM

I would think that as long as the voltage from the VRM is above the chip's lower operating threshold, it should not make any difference, unless they are being overclocked heavily.

The temp inversion article has me sold on it that 28nm chips run better a bit warm
sr. member
Activity: 336
Merit: 250
October 26, 2013, 12:17:54 PM
Thought I would share this and maybe some here can even update it with new retailers....

http://coinmap.org/
legendary
Activity: 1066
Merit: 1098
October 26, 2013, 12:16:45 PM
this "higher temp =  higher performance" doesnt make a lot of sense. Design aside, there are only two factors affecting performance:
- clockspeed. Is this variable on KnC? Can it be measured?
- number of functional cores/ % of HW errors. Does this go down as temps go up?

Thats it. The rest is statistical noise.

sense or not. i have one jupiter that reacts in the same way. additional cooling with a big box-fan on top of the miner = cgminer crashes every 5 min. without box-fan = cgminer stable / higher avg GH.

Can it be electrical interference from the fan's motor? They have quite a strong electromagnetic field when running.

It's probably a effect called temperature inversion. 28nm logic gets faster with higher temperatures.
http://siliconsaint.blogspot.com/2012/07/temperature-inversion-in-deep-sub.html?m=1

I don't think so...  That phenomenon might make it possible to clock a chip faster at a higher temperature, but it should not effect the speed of a chip that is clocked at a constant speed.  Right?



I honestly think it's a VRM thing... not an asic thing..  Smiley

I would think that as long as the voltage from the VRM is above the chip's lower operating threshold, it should not make any difference, unless they are being overclocked heavily.
Jump to: