[Guide] Dogie's Comprehensive ASICMiner Cube Setup [HD]

[sidehack]

More than likely it's the burst current when the full-power relay clicks on and sources current to the cards. If they staged power-on would be easier, but these things go from zero to max basically instantly. Some power supplies don't have enough output capacitance to hold that current without undervolting the supply and tripping it off.

[lajz99]

Has anyone taken their out of the case in order to keep it running cooler?

[babaji.ca]

More than likely it's the burst current when the full-power relay clicks on and sources current to the cards. If they staged power-on would be easier, but these things go from zero to max basically instantly. Some power supplies don't have enough output capacitance to hold that current without undervolting the supply and tripping it off.

yes, i think that may be it, sounds familiar. the guy found a solution by wiring something up to hack it. I wish i could find that post again!

[babaji.ca]

Dogie and sidehack:

After power on (via paperclip) the cubes (at low clock) turn on but then when the relays click (a second or so later) the PSU shuts down. Removing the paperclip and adding it back results in the exact same behavior

I actually had this issue with a hacked-up 500W server supply. The problem was, when the relays trip it requires a large burst current that drains the output capacitors inside the power supply. I tucked a couple 3300uF 16V caps into the 6-pin connectors, such that they would charge up when the power supply kicked on, and then supply burst current requirements when the miner fully fired up. This reduces the burst load on the power supply, which may be tripping overcurrent or undervolt protection (as the output cap voltage drops out before the regulator can compensate).

Find a couple good-sized 16+V capacitors at Radio Shack or something and wire them onto the 12V lines and it'll probably solve the problem.

Perfect. I had a couple of 1000uF 16V caps that were enough to get the relays jump started w/o tripping the PSU protection. I just temporarily wedged them into the connector to get it started (then I removed them once the relays clicked). I'll come up with a more elegant sol'n later. Thanks again.

-dmc

Hi, I am having this issue also. Do you have any pics of this to help me? Did you find a more permanent solution?

[lajz99]

Has anyone taken their out of the case in order to keep it running cooler?

I took mine out of the case and I'm getting ~400mh/s per cube increased.  The board with the heatsink nearest to the case has numbers on par with the rest of the boards now.   Does anyone else notice their end board hashes slower?

[dresdenreader]

Dogie, just wanted to say, thanks for the guide! Got my cube today and got it up and running after about 2-3 hours of troubleshooting. The router gave me the biggest amount of trouble but after that was fixed I was up and running at around 38GH. If it wouldn't have been for this guide I definitely wouldn't have got it turned on this fast.

Thanks!

[dresdenreader]

Also, new question.

Now that I've got the cube running, what is the suggested mining difficulty? I've got it at 64 right now with according to the cube page around 96% efficiency.

[babaji.ca]

Also, new question.

Now that I've got the cube running, what is the suggested mining difficulty? I've got it at 64 right now with according to the cube page around 96% efficiency.

Glad to hear you got yours working!

I have only got one going now. Running it on an antec 380 watt from my HTCP in low mode. Turns out my mega 1050 PSU isnt such a tough guy after all. (he has issues)

one out of the three I got today had a screw rattling around inside. It was from one of the heat sinks. The other three screws on that heat sink was 50% of the way unscrewed and was so loose that the heat sink slapped around as I rotated  the cube in my hand. Thankfully she fixed up nice.

Now I just need to get some juice for the other two idle cubes.

What did you end up doing for power? Are you making do with the 2 PSU you had kicking around?

[dresdenreader]

Well, I've only got one cube so far, ordered from SSB and Canary. I ordered the Cube from SSB after ordering from Canary and the SSB Cube got here first. Haven't heard word from Canary about the cubes, but often with him no news is good news.

I'm currently using the Antec 620 that I had around and I've had zero problems, now to see if it can last through the night. I've got the Corsair 800 GS on order from Amazon that will be here Friday. I'm fairly confident the PSU will get here before the other two cubes do sadly.

What worker difficulty are you running your one at? Also what pool?

[babaji.ca]

I am hashing on Eligius. Not sure what diff i am running em at? Stock? I guess. Where do you see that setting?

anyways in low mode it shows  31500 or so MHS on the config page and 95.28% efficiency

[doltek]

I have one arriving in a few hrs. How do i config with ghash.io?

Sorry, Im kinda a n00b

[dmcdad]

Dogie and sidehack:

After power on (via paperclip) the cubes (at low clock) turn on but then when the relays click (a second or so later) the PSU shuts down. Removing the paperclip and adding it back results in the exact same behavior

I actually had this issue with a hacked-up 500W server supply. The problem was, when the relays trip it requires a large burst current that drains the output capacitors inside the power supply. I tucked a couple 3300uF 16V caps into the 6-pin connectors, such that they would charge up when the power supply kicked on, and then supply burst current requirements when the miner fully fired up. This reduces the burst load on the power supply, which may be tripping overcurrent or undervolt protection (as the output cap voltage drops out before the regulator can compensate).

Find a couple good-sized 16+V capacitors at Radio Shack or something and wire them onto the 12V lines and it'll probably solve the problem.

Perfect. I had a couple of 1000uF 16V caps that were enough to get the relays jump started w/o tripping the PSU protection. I just temporarily wedged them into the connector to get it started (then I removed them once the relays clicked). I'll come up with a more elegant sol'n later. Thanks again.

-dmc

Hi, I am having this issue also. Do you have any pics of this to help me? Did you find a more permanent solution?

I don't have any photos but the solution was pretty inelegant. I just shoved the capacitor leads into the back of the power connectors going into the cubes (I used 2 capacitors on each cube). Just make sure the negative lead on the capacitor goes to the black/neutral wired in the connector. After it started up I just removed the capacitors and set them down next to the cubes in case they lose power and I have to restart them. I was going to come up with a more permanent solution but I haven't had to power down the cubes yet so the motivation is quite low.

-dmc

[novymivo]

More than likely it's the burst current when the full-power relay clicks on and sources current to the cards. If they staged power-on would be easier, but these things go from zero to max basically instantly. Some power supplies don't have enough output capacitance to hold that current without undervolting the supply and tripping it off.

yes, i think that may be it, sounds familiar. the guy found a solution by wiring something up to hack it. I wish i could find that post again!

mate
try to find this for us.
I can add , when the clicking noise starts, the bottom gren light start flashing as well.

[sidehack]

I don't have any photos but the solution was pretty inelegant. I just shoved the capacitor leads into the back of the power connectors going into the cubes (I used 2 capacitors on each cube). Just make sure the negative lead on the capacitor goes to the black/neutral wired in the connector.

That's exactly how I did it also.

[babaji.ca]

Thanks for the description. Could you please see if i got this straight.

1 cap per pcie connector (ie one on the left connector and one on the right connector, 2 per cube total)


I am kind of new to this. There are two(2) rows of (3) pins on each pcie connector. mine had a row of 3 yellow and a row of 3 black pins .

Black is ground? (it's the row adjacent the locking clip mechanism) So I put the negative lead of the cap into the black row. And the positive lead into the yellow row. Does it matter which column I insert them into. Ie position 1, 2, or 3?

[babaji.ca]

I have one arriving in a few hrs. How do i config with ghash.io?

Sorry, Im kinda a n00b

I'll take a stab.

The guide is pretty good. Follow it. The only unclear thing I can help you with (maybe). Is the stratum server you need to set up.  In linux U run it as follows:  

./mining_proxy.py -o stratum.mining.eligius.st -p 3334

notice the

-o <your pools address>  

and

-p <port number>

I am not sure if I am to have my local stratum server connecting to eligius using stratum? but it seems to be working well.

locally here on my net, the cubes are connecting to my stratum server using getwork. again , not sure if that is ideal. but it seems to work well for me.

....hopefully someone who knows what they are doing can chime in.

[sidehack]

The leftmost pair on the left connector is your standby power, so capping that won't do much. But any of the rest will route straight into your cards' power inputs. I'd just stick it in the middle of each one for simplicity. Yes, the black is ground which should go to the marked lead on the cap, which should be the shorter of the two leads. The other lead goes to yellow. I'd recommend at least 1000uF 16V if you got 'em, and actually trim the leads down about as short as will still make good contact. Longer leads mean resistance and inductance, neither of which are, in this case, very extreme but will still marginally hurt burst-current performance.

[babaji.ca]

Thanks for the explanation. A big help.

Do I have to install them physically  close to the cube? could I splice them into an extension cable? (maybe midway down a 6 inch cable)? 

Thanks!

[babaji.ca]

just <FYI>

installed stratum to ras pi. The python process took up 85-95% of cpu with one cube on low. Ran well 31,000 mhs. however the second cube would not get above 15,000 mhs. I think the raspi was saturated. moved server duty to a real machine, works well now. getting full speed from two cubes while the third is still in low mode due to toy PSU.

bought another powersupply this morning, a thermaltake  850w bronze, it has two cubes in high mode. seems to draw about 770 W according to kill-a-watt meter. a little more than advertised i guess. Not sure if that is accurate.

I better get the seasonic 1050 watt back in service soon!

[Jobe7]

Guys my miner was working at 38gh fine, now its back down to 30? Any reason why and how can i increase back to 38 again?