Avalon ASIC users thread

[driksson]

Hi,

batch2 does not have proper chip heat reading? so the 20, 40, 42 reading in reality is 40, 60 and 62 reading?
It ran fine for 3 days straight, then it just went to zero hashing.. and restarting it gets back to hashing a while, then goes down again.
--avalon-auto freq 300-355
what is best fan setting? it is in a 20dgr ac room..

Applying changes
Loading /etc/config/cgminer

its just stuck there! just reflashed to factory defaults, it gets an ip from the wifi.
[Firmware Version] => 20130821
cgminer-70bc62c
luci-d854edc+
cgminer-openwrt-packages-800cc58
Socket connect failed: Connection refused

thats what cgminer api says.
ive tried waiting for 2 minutes.. and restart. it was working fine for 3 days straight.
should i try an older firmware?

[GandalfG]

Hi,

batch2 does not have proper chip heat reading? so the 20, 40, 42 reading in reality is 40, 60 and 62 reading?
It ran fine for 3 days straight, then it just went to zero hashing.. and restarting it gets back to hashing a while, then goes down again.
--avalon-auto freq 300-355
what is best fan setting? it is in a 20dgr ac room..

Applying changes
Loading /etc/config/cgminer

its just stuck there! just reflashed to factory defaults, it gets an ip from the wifi.
[Firmware Version] => 20130821
cgminer-70bc62c
luci-d854edc+
cgminer-openwrt-packages-800cc58
Socket connect failed: Connection refused

thats what cgminer api says.
ive tried waiting for 2 minutes.. and restart. it was working fine for 3 days straight.
should i try an older firmware?

Try this
https://dl.dropboxusercontent.com/u/16881231/avalon_clone/firmware/openwrt-ar71xx-generic-tl-wr703n-v1-squashfs-factory_20130519b.bin
For testing and  OC best one.

[driksson]

Hi,

batch2 does not have proper chip heat reading? so the 20, 40, 42 reading in reality is 40, 60 and 62 reading?
It ran fine for 3 days straight, then it just went to zero hashing.. and restarting it gets back to hashing a while, then goes down again.
--avalon-auto freq 300-355
what is best fan setting? it is in a 20dgr ac room..

Applying changes
Loading /etc/config/cgminer

its just stuck there! just reflashed to factory defaults, it gets an ip from the wifi.
[Firmware Version] => 20130821
cgminer-70bc62c
luci-d854edc+
cgminer-openwrt-packages-800cc58
Socket connect failed: Connection refused

thats what cgminer api says.
ive tried waiting for 2 minutes.. and restart. it was working fine for 3 days straight.
should i try an older firmware?

Try this
https://dl.dropboxusercontent.com/u/16881231/avalon_clone/firmware/openwrt-ar71xx-generic-tl-wr703n-v1-squashfs-factory_20130519b.bin
For testing and  OC best one.

Thanks, i will try it later. I even tried standard chip freq 280, and without --avalon-auto..
Just doesnt seem to connect to cgminer?

[driksson]

Hi,

batch2 does not have proper chip heat reading? so the 20, 40, 42 reading in reality is 40, 60 and 62 reading?
It ran fine for 3 days straight, then it just went to zero hashing.. and restarting it gets back to hashing a while, then goes down again.
--avalon-auto freq 300-355
what is best fan setting? it is in a 20dgr ac room..

Applying changes
Loading /etc/config/cgminer

its just stuck there! just reflashed to factory defaults, it gets an ip from the wifi.
[Firmware Version] => 20130821
cgminer-70bc62c
luci-d854edc+
cgminer-openwrt-packages-800cc58
Socket connect failed: Connection refused

thats what cgminer api says.
ive tried waiting for 2 minutes.. and restart. it was working fine for 3 days straight.
should i try an older firmware?

Try this
https://dl.dropboxusercontent.com/u/16881231/avalon_clone/firmware/openwrt-ar71xx-generic-tl-wr703n-v1-squashfs-factory_20130519b.bin
For testing and  OC best one.

After a few hours off, it connected cgminer (i did no changes of firmware, i still run the 0821 version). i now have it on std freq 300, avalon-auto 300-350 avalon-fan 80 , or should i do 100 on fan? its damn noisy..
heat says 30,40,40, so i believe its 40,60,60? ill let it run a few hours, and then check heat again and maybe put fan on 90?
It was on for a while, then it stopped hashing again.. i am fearing that its the wifi thats not connecting too well. or is it really overheating with 300-350 on fan at 80%?

[-ck]

After a few hours off, it connected cgminer (i did no changes of firmware, i still run the 0821 version). i now have it on std freq 300, avalon-auto 300-350 avalon-fan 80 , or should i do 100 on fan? its damn noisy..
heat says 30,40,40, so i believe its 40,60,60? ill let it run a few hours, and then check heat again and maybe put fan on 90?
It was on for a while, then it stopped hashing again.. i am fearing that its the wifi thats not connecting too well. or is it really overheating with 300-350 on fan at 80%?

You got it backwards. Temp readings of up to 60 are fine on batch 1 and 2. Temp readings up to 80 are fine on batch 1 and 2. I can't figure out which firmware you ended up using, but mine will just adjust the fanspeed automatically fine for temperature and you shouldn't set any temps or fanspeeds if you have a batch 1 or 2.

[tarui]

try setting min fan speeds

i found a potential problem with the 20120721, not sure if it's fixed as i havent tried other versions yet

the funny thing is, once temps go below 48 (i set the threshold at 52) the fans drop to min rpm 840rpm then temp rises quickly, sends the fans spinning like mad again then cycle repeats

[SolarSilver]

Temp readings of up to 60 are fine on batch 1 and 2. Temp readings up to 80 are fine on batch 1 and 2.

Don't you mean

Temp readings up to 80 are fine on batch 3.

?

[zoro]

yes 80 for B3
but i thought than for b1&2 temps above 50 are dengerous for 24/7 operation. am i wrong? can i have 55 or even 59 for 24/7?

[loshia]

What value resistor would I piggyback onto R10 to target 1.30V, or 1.25V?

For 1.3V (1.278V under load) I use 47k. No problem with 400MHz and 0.98% HW, Rejected 0.38%
For stability I suggest sold 2mm² wire from coil to the end of board. Top cooper layer (VDD_1.2)  is too thin to carry such high currents.
Look on the picture below
https://www.dropbox.com/s/68f6k5tafw1i58f/20130825_025617-2.jpg

Im running with 31k piggyback and that gives me 1.336v under load at the bottom chip.

Be aware that for me at 460mhz I am 2x the wattage as a stock Avalon!   At bare MINIMUM you need a 1200watt powersupply for a 4 module unit!

I need to find out how to deal with 1600watt total at 460mhz!!!

Guy's,

I am wandering how you are able to get such higher rates? I am not hardware expert but according to my knowledge the TPS DC/DC converter can handle up to 20 Amps. At clocks higher that 400 maybe chip consumes 3-4A at 1.3V so 10 Chips *4A=40 Amps. Double the capacity that  TPS DC/DC can handle.

I would be more than happy if someone comments it
Thanks

[driksson]

After a few hours off, it connected cgminer (i did no changes of firmware, i still run the 0821 version). i now have it on std freq 300, avalon-auto 300-350 avalon-fan 80 , or should i do 100 on fan? its damn noisy..
heat says 30,40,40, so i believe its 40,60,60? ill let it run a few hours, and then check heat again and maybe put fan on 90?
It was on for a while, then it stopped hashing again.. i am fearing that its the wifi thats not connecting too well. or is it really overheating with 300-350 on fan at 80%?

You got it backwards. Temp readings of up to 60 are fine on batch 1 and 2. Temp readings up to 80 are fine on batch 1 and 2. I can't figure out which firmware you ended up using, but mine will just adjust the fanspeed automatically fine for temperature and you shouldn't set any temps or fanspeeds if you have a batch 1 or 2.

I will leave all temp settings out then. However, what could be the reason for it to just stop mining? fans are not going high at all.
Could it be connectivity that dies? Its on wifi for now.
Thank you for your time!

[driksson]

yes 80 for B3
but i thought than for b1&2 temps above 50 are dengerous for 24/7 operation. am i wrong? can i have 55 or even 59 for 24/7?

I'd imagine that also. since b2 readings are not on chip and should be 20dgrs cooler than chip. at 50+ your chip reading should be 70+.

[GenTarkin]

What value resistor would I piggyback onto R10 to target 1.30V, or 1.25V?

For 1.3V (1.278V under load) I use 47k. No problem with 400MHz and 0.98% HW, Rejected 0.38%
For stability I suggest sold 2mm² wire from coil to the end of board. Top cooper layer (VDD_1.2)  is too thin to carry such high currents.
Look on the picture below
https://www.dropbox.com/s/68f6k5tafw1i58f/20130825_025617-2.jpg

Im running with 31k piggyback and that gives me 1.336v under load at the bottom chip.

Be aware that for me at 460mhz I am 2x the wattage as a stock Avalon!   At bare MINIMUM you need a 1200watt powersupply for a 4 module unit!

I need to find out how to deal with 1600watt total at 460mhz!!!

Guy's,

I am wandering how you are able to get such higher rates? I am not hardware expert but according to my knowledge the TPS DC/DC converter can handle up to 20 Amps. At clocks higher that 400 maybe chip consumes 3-4A at 1.3V so 10 Chips *4A=40 Amps. Double the capacity that  TPS DC/DC can handle.

I would be more than happy if someone comments it
Thanks

According to the guys that have done the mentioned mods, thats a soft limit of the DC/DC converter. Meaning exactly what, I dont know =P ... I wonder if it has to do w/ hard limit of maximum POWER(watts) output as in maybe its max capable voltage @ 20amps is what its rated for. Which is a lot more than the voltage we put through it.

[S M I L Y]

Guy's,

I am wandering how you are able to get such higher rates? I am not hardware expert but according to my knowledge the TPS DC/DC converter can handle up to 20 Amps. At clocks higher that 400 maybe chip consumes 3-4A at 1.3V so 10 Chips *4A=40 Amps. Double the capacity that  TPS DC/DC can handle.

I would be more than happy if someone comments it
Thanks

There are really only 2 areas where a switching VR can fail.  Heat and inductor saturation.     If the FET cant handle the current the friction of the electrons traveling through the junction turns to heat.  The heat will eventually kill the part.   If you keep things cool (im watercooling) you can go amazingly far past design tolerances.

I dont know what the inductor saturation point is.  I havent looked up the spec.  I can say that I can hold my finger directly on the inductor and it dosnt burn me.  Its not comfortable, probably about 45-50c temp.

As far as killing my Avalon goes... I dont care.  I need to get as much out of this thing in the next 30 days... after that it probably wont be profitable to run.

[driksson]

i think i found the cause of my problem, i read somewhere, that the current is not strong enough to the wifi unit, and since i dont have an antennae, its around 30% reception only. this makes the connection strong in the beginning (before overclocking kicks in), and when OC is on max, there is not enough for the wifi, and then it starts slowly dying off.. it takes around 30min to one hour to go from 100% hash speed down to basically thousands of hashes.. will try lan tomorrow. but annoying as it worked just fine for three days!

[loshia]

Guy's,

I am wandering how you are able to get such higher rates? I am not hardware expert but according to my knowledge the TPS DC/DC converter can handle up to 20 Amps. At clocks higher that 400 maybe chip consumes 3-4A at 1.3V so 10 Chips *4A=40 Amps. Double the capacity that  TPS DC/DC can handle.

I would be more than happy if someone comments it
Thanks

There are really only 2 areas where a switching VR can fail.  Heat and inductor saturation.     If the FET cant handle the current the friction of the electrons traveling through the junction turns to heat.  The heat will eventually kill the part.   If you keep things cool (im watercooling) you can go amazingly far past design tolerances.

I dont know what the inductor saturation point is.  I havent looked up the spec.  I can say that I can hold my finger directly on the inductor and it dosnt burn me.  Its not comfortable, probably about 45-50c temp.

As far as killing my Avalon goes... I dont care.  I need to get as much out of this thing in the next 30 days... after that it probably wont be profitable to run.

Thank you for your prompt response.
If avalon will not be profitable in a month i am curious what will be profitable at all?

The money printing machines are history...And we have to get used to it.


Best

[GandalfG]

Guy's,

I am wandering how you are able to get such higher rates? I am not hardware expert but according to my knowledge the TPS DC/DC converter can handle up to 20 Amps. At clocks higher that 400 maybe chip consumes 3-4A at 1.3V so 10 Chips *4A=40 Amps. Double the capacity that  TPS DC/DC can handle.

I would be more than happy if someone comments it
Thanks

There are really only 2 areas where a switching VR can fail.  Heat and inductor saturation.     If the FET cant handle the current the friction of the electrons traveling through the junction turns to heat.  The heat will eventually kill the part.   If you keep things cool (im watercooling) you can go amazingly far past design tolerances.

I dont know what the inductor saturation point is.  I havent looked up the spec.  I can say that I can hold my finger directly on the inductor and it dosnt burn me.  Its not comfortable, probably about 45-50c temp.

As far as killing my Avalon goes... I dont care.  I need to get as much out of this thing in the next 30 days... after that it probably wont be profitable to run.

TPS chip are designed to handling 20A but in reality all depends how much can drive  MOSFET. If FET can switch more amps and have internal resistance and capacity below the level to which it has been TPS chip designed for we are safe.
It is more important to eliminate bottlenecks like resistance of the transistor and the current paths. That's  unnecessarily generate additional heat.

[greaterninja]

2 weeks ago my Batch #3, 4 module stopped hashing using a 850w power supply.

I had to disconnect the motherboard atx power connector to get it back online.
When I reconnected it a huge bright spark came from module #4.


Now 2 weeks later, using a 1050w gold corsair, my 4 module Batch #3 has stopped hashing again.

I disconnect the atx motherboard connector and then reconnect it.  Avalon powers back on, however module #2 has a huge spark come from it as well.   It looks like a huge glow from a capacitor each time.


Is this something to be concerned about?   Has anyone else had this happen to them?  The unit is hashing fine now...I just wonder what is happening.


I have the following connectors hooked up:

1- atx motherboard
2- pci-e 8 pin connectors
2- CPU 8 pin connectors

Each time this has happened, I have had to reset power to my controller board to get the thing mining properly again.

[GandalfG]

Can you post photo from broken hash module ?

[loshia]

Guy's,

I am wandering how you are able to get such higher rates? I am not hardware expert but according to my knowledge the TPS DC/DC converter can handle up to 20 Amps. At clocks higher that 400 maybe chip consumes 3-4A at 1.3V so 10 Chips *4A=40 Amps. Double the capacity that  TPS DC/DC can handle.

I would be more than happy if someone comments it
Thanks

There are really only 2 areas where a switching VR can fail.  Heat and inductor saturation.     If the FET cant handle the current the friction of the electrons traveling through the junction turns to heat.  The heat will eventually kill the part.   If you keep things cool (im watercooling) you can go amazingly far past design tolerances.

I dont know what the inductor saturation point is.  I havent looked up the spec.  I can say that I can hold my finger directly on the inductor and it dosnt burn me.  Its not comfortable, probably about 45-50c temp.

As far as killing my Avalon goes... I dont care.  I need to get as much out of this thing in the next 30 days... after that it probably wont be profitable to run.

TPS chip are designed to handling 20A but in reality all depends how much can drive  MOSFET. If FET can switch more amps and have internal resistance and capacity below the level to which it has been TPS chip designed for we are safe.
It is more important to eliminate bottlenecks like resistance of the transistor and the current paths. That's  unnecessarily generate additional heat.

10X dude
i have looked at your photo with wires:) and i liked it very much. that explains things for us none experts.

have yo measured how much a module/chip draws at 400+ with avalon design?

[-ck]

Temp readings of up to 60 are fine on batch 1 and 2. Temp readings up to 80 are fine on batch 1 and 2.

Don't you mean

Temp readings up to 80 are fine on batch 3.

?

Lol yeah. Lacking sleep. When I say fine up to 60 degrees, I'm not recommending running them at 60. At the risk of repeating myself a million times over - the defaults are in the code for a reason. By default my firmware will try to keep the temperature under 50.