Avalon ASIC users thread

[allinvain]

well, there's definitely a difference of opinion around here in regards to UPS.

both galambo and Death & Taxes have gone on record against UPS with these things.  you'd have to hunt about the details though in other threads.

Well I think the main problem with putting a UPS behind these things is that they're absolute power consumption monsters and it takes one beefy ass UPS to feed them for a reasonable amount of time - enough to last the average blackout.

I was thinking of getting some 1440 or 1500VA units. Haven't pulled the trigger yet though. Where did galambo and Death & Taxes discuss this?

here's D&T's response when i asked him this in a pm:

"Getting a UPS capable of handling that kind of wattage is going to be expensive plus all UPS add load and inefficiency.  If it were me I wouldn't be considering a UPS. "

Yep I agree completely with him. The only reason I guess I even considered getting a few USPs is paranoia. Maybe hosting these at a data center is the best solution (if you have multiple as that would justify the potentially high monthly cost).

[cypherdoc]

Also in regards to the avalon unit coming back online from a power outage, sometimes that does not happen. Rarely what could happen is the unit will be left in a zombie like state because the control board did not initialize properly. For me this happened once when there was a brownout.

would the control board respond to a ping in that situation from this device?  http://dataprobe.com/iboot-g2.php

Bear in mind that router board (which runs the cgminer software and management front end) is NOT the control board. What I mean by the control board is the Spart6 FPGA board that takes the instructions from the cgminer software running on the router board and coverts that to work batches for the ASICS. The control board is the one with all the "IDE like" cables running out of it and to the individual hashing modules. The control board does not respond to ping requests because it is not accessible via any network protocols as it's attached to the router board via USB. The router board responds to ping request and I'd imagine it would also respond to ping request even if the mining software is not running or in some sort of locked up or crashed state. If the router's OS was completely down or broken then yes it would not respond to ping requests. The power control devices - dataprobe - are of limited use in this scenario. They are still useful for when you want to quickly power cycle the machine for whatever reason.

It would be neat if they made a 2 socket version of the iBoot-G2. The reason for 2 is that ideally you'd want 2 Avalon units per 15 Amp circuit. The one with multiple power sockets has only 1 power in cable.

Edit: Ooops..seems there is a iBoot-G2+ which allows you to control up to 3 power outlets. Neat!

Yep, it is a neat device but you're stuck having to constantly ssh in to try and catch the rare zombie like condition.

What other solutions are there assuming they're kept off site?

[opentoe]

I highly recommend Seasonic PSU's. I've used them in all my computers and they are of awesome quality and very efficient. Check out their X series.

http://www.seasonicusa.com/NEW_X-series_1050-1250.htm

[yxt]

but seasonic are (too?) large

[Sitarow]

I am not sure if this has been answered however.

Is there some way to increase the fan speed on the units?

[cypherdoc]

Also in regards to the avalon unit coming back online from a power outage, sometimes that does not happen. Rarely what could happen is the unit will be left in a zombie like state because the control board did not initialize properly. For me this happened once when there was a brownout.

would the control board respond to a ping in that situation from this device?  http://dataprobe.com/iboot-g2.php

Bear in mind that router board (which runs the cgminer software and management front end) is NOT the control board. What I mean by the control board is the Spart6 FPGA board that takes the instructions from the cgminer software running on the router board and coverts that to work batches for the ASICS. The control board is the one with all the "IDE like" cables running out of it and to the individual hashing modules. The control board does not respond to ping requests because it is not accessible via any network protocols as it's attached to the router board via USB. The router board responds to ping request and I'd imagine it would also respond to ping request even if the mining software is not running or in some sort of locked up or crashed state. If the router's OS was completely down or broken then yes it would not respond to ping requests. The power control devices - dataprobe - are of limited use in this scenario. They are still useful for when you want to quickly power cycle the machine for whatever reason.

It would be neat if they made a 2 socket version of the iBoot-G2. The reason for 2 is that ideally you'd want 2 Avalon units per 15 Amp circuit. The one with multiple power sockets has only 1 power in cable.

Edit: Ooops..seems there is a iBoot-G2+ which allows you to control up to 3 power outlets. Neat!

Yep, it is a neat device but you're stuck having to constantly ssh in to try and catch the rare zombie like condition.

What other solutions are there assuming they're kept off site?

since nobody seems to have the answer to this question, i'll just ask another.

the iBoot has a Windows software protocol called Heartbeat which can be installed on a computer which will send "heartbeat" signals to the iBoot periodically to indicate its functioning.  this is the reverse direction of the ping function originating from the iBoot.

i wonder if it would be possible to write a linux version of this and install it into openwrt and have it monitor the control board and make it fail to send "heartbeats" to iBoot in the case of a control board freeze which would then trigger a reboot?

[allinvain]

Yep that should be possible to do. However I hear that the dev crew are labouring to come up with with a workaround for the control board issues.

[ewibit]

I have changed the IP of Avalon from 192.168.0.100 to 192.168.1.100 (255.255.0.0) to better fit to the router IP
now I cannot access to luci anymore across Webbrowser or ssh (WLAN)
(in router there is luci/Avalon with the IP 192.168.1.x but ping does not work

Code:

PING 192.168.1.x (192.168.1.x) 56(84) bytes of data

is it ok that it seems that IP of Avalon LAN is the same as Avalon WLAN?
I have not located anywhere to change the WLAN IP, only the LAN IP...

[allinvain]

I have changed the IP of Avalon from 192.168.0.100 to 192.168.1.100 (255.255.0.0) to better fit to the router IP
now I cannot access to luci anymore across Webbrowser or ssh (WLAN)
(in router there is luci/Avalon with the IP 192.168.1.x but ping does not work

Code:

PING 192.168.1.x (192.168.1.x) 56(84) bytes of data

is it ok that it seems that IP of Avalon LAN is the same as Avalon WLAN?
I have not located anywhere to change the WLAN IP, only the LAN IP...

Go to the network tab. It's all there. Edit the WLAN interface and give it a different IP otherwise you are going to continue to have IP conflicts.

[irritant]

I have changed the IP of Avalon from 192.168.0.100 to 192.168.1.100 (255.255.0.0) to better fit to the router IP
now I cannot access to luci anymore across Webbrowser or ssh (WLAN)
(in router there is luci/Avalon with the IP 192.168.1.x but ping does not work

Code:

PING 192.168.1.x (192.168.1.x) 56(84) bytes of data

is it ok that it seems that IP of Avalon LAN is the same as Avalon WLAN?
I have not located anywhere to change the WLAN IP, only the LAN IP...

Go to the network tab. It's all there. Edit the WLAN interface and give it a different IP otherwise you are going to continue to have IP conflicts.

I guess he can't access the network tab, maybe if you connect it to your laptop with ethernet cable and and assign static ip to your laptop (192.168.0.101)  you might be able to access luci across webbrowser again ?

[allinvain]

Yeah sorry I forgot to mention to try connecting via Ethernet first. I assumed he would know that though. If that fails he can always reset the system to factory defaults: http://wiki.openwrt.org/toh/tp-link/tl-wr703n

But I'd recommend trying a straight Ethernet connection before messing with the reset button.

[kaerf]

FYI, this morning my Avalon stopped hashing. It was hashing fine before I went to sleep, but after restarting cgminer and issuing the "reboot" command at the openwrt prompt, it still didn't work. I had to manually power cycle the machine by turning off the power to get the machine hashing again. I was using the latest 20130410 firmware. Gave me a good scare.

Here's the error that cgminer was giving:
AVA0: FPGA controller mess up, 24 wrong results                   
Avalon: Reset failed! not an Avalon? (0: c1 4f 87 64)

[allinvain]

FYI, this morning my Avalon stopped hashing. It was hashing fine before I went to sleep, but after restarting cgminer and issuing the "reboot" command at the openwrt prompt, it still didn't work. I had to manually power cycle the machine by turning off the power to get the machine hashing again. I was using the latest 20130410 firmware. Gave me a good scare.

Here's the error that cgminer was giving:
AVA0: FPGA controller mess up, 24 wrong results                   
Avalon: Reset failed! not an Avalon? (0: c1 4f 87 64)

Yeah...umm...there is an issue with the FPGA controller board. I heard on the grapevine that they are working on a firmware workaround for it. Stay tuned to this thread.

Btw, what was your uptime before the lockup?

[loshia]

Yeah...umm...there is an issue with the FPGA controller board. I heard on the grapevine that they are working on a firmware workaround for it. Stay tuned to this thread.

Btw, what was your uptime before the lockup?

You mean a new bitstream for spartan 6?  That will be great! Are you sure about it dude? Who spreads this rumor..?

[Aseras]

cant fix the fpga, its a bug in hardware ( according to xiangfu ). We are going to ditch the serial code and use direct usb. new drivers.  ckolivas and xiangfu are going to try and do that next week. That should let us work around the problem and make things much more fluid.

[loshia]

cant fix the fpga, its a bug in hardware ( according to xiangfu ). We are going to ditch the serial code and use direct usb. new drivers.  ckolivas and xiangfu are going to try and do that next week. That should let us work around the problem and make things much more fluid.

Super!
10X for the update!

[Mikej0h]

What are safe temperature levels for the Avalon units?
I'm looking currently at T1: 27, T2: -1, T3: 46
Where T3 fluctuates between 44 and 47.

Also I'm having for ~20 hours of mining 1150 HW errors, is this usual?

The status screen anyways brings up alot of questions, in sence of; what do all the abriations mean (didn't find it on the wiki, and I'm pretty new to mining).

[opentoe]

but seasonic are (too?) large

They may be physically too large...have to get exact dimensions I guess. That's probably the one thing you don't want to be cheap on, is a good PSU and I think Seasonic's are the best in the business.

[opentoe]

Since this "is" Avalon users thread let us poor people know how much money you guys have made that you have CASHED in, NOT how many bitcoins you have accumulated. I'm quite curious. Of course if you want to refrain from letting us know how much money you have made that is your liberty and don't need any nasty offending comments that I've offended you for asking.

I just want to sulk and go into a deep depression from hearing all the numbers.

[luffy]

Since this "is" Avalon users thread let us poor people know how much money you guys have made that you have CASHED in, NOT how many bitcoins you have accumulated. I'm quite curious. Of course if you want to refrain from letting us know how much money you have made that is your liberty and don't need any nasty offending comments that I've offended you for asking.

I just want to sulk and go into a deep depression from hearing all the numbers.

you ask for it :p
20k$ in a month, my year's salary