Topic: ANTMINER U3 Discussion and Support Thread - page 52. (Read 149493 times)

Good to know. I did check make sure it wasn't touching the pins.
Very interesting white paper.
So what do you recommend then to be used generally?

From my "CPU Thermal Pasting" days, I thought the whole idea was to use the stuff VERY sparingly, even more so than the very small "new blobs" shown. In the CPU days it almost seemed like an art form to get the amount "just right", certainly nothing like what was done at the factory.

Yep - that's the stuff.

Do you mean arctic ceramique? That's specifically designed for open chip designs [ie GPUs, miners] as its nearly non conductive.

Funny... I used AS (white version) and it seemed to "fix" my zombie issue.

Perhaps it's more a matter of how well the compound is applied versus its conductivity.

You should NOT use AS5 on a unit like this which has exposed pins and other electronics on the PCB. AS5 is conductive and may cause problems. Additionally, its thermal conductivity is only a 1/10th of what they claim it to be, is 4x less than other competitors and up to 5x more expensive. You can see the full independent paper here, results on page 8.

Yep, still looking for someone who has an unmodified miner that will go more than 24hrs without going ZOMBIE.

Dogie, you said there are 1000's of U3 owners out there with good units, where are they?

When it zombies, let us know... lol 

Well after all the info in this thread I am up and running. Thank you to all that contributed.

Setup:
Laptop running Fedora 20
1x U3
Power brick 19v @ 3.46A

Running around 60GH/s

HAHAHAHAHA. That made me LOL. I used the same stuff... Alcohol wipes but then used a toothpick and toothbrush to really get the crap off the pcb and smd's...  Hope it solves your problem, excited to see.

OK I opened mine up and cleaned all that paste out, I forgot the stuff between the heatsink and the PCB..... shouldnt matter that much on that side.

This thermal paste is like stuffed inbetween the pins of the ASIC's... I couldnt clean it out without getting more aggressive and the last thing I needed is to kill an ASIC so I did my best by gently cleaning it for an hour and using alco wipes and some other SMD tools.

https://i.imgur.com/DiLIz5O.jpg

Fused spunk allover the fucking place

https://i.imgur.com/3LQeHyv.jpg

https://i.imgur.com/cl1lOOr.jpg

https://i.imgur.com/SerSo5j.jpg

Arctic Silver (Silver Oxide thermal paste)

https://i.imgur.com/yyQeqts.jpg

Chilling at 62.5ghs on top of his brother

https://i.imgur.com/k4LNjLt.jpg


Ill let it run till tomorrow. We see what happens.

So check it out... Looks like I have been up and running for over 39 hours with no zombie...  However, it looks like only 3 of the 4 chips are working on my unit; thus the lower hashrates...

39 Hours before exiting:
https://i.imgur.com/XEbnJVB.jpg

Receive 1-3, missing #4:
https://i.imgur.com/RNlWvYI.jpg

WOW. I should pop mine open to see what it looks like too...  Yet folks seem to think that they only melt at stock settings...  They must have received a nicer version than us...

More evidence of some problems with the supplied power supply. This happened on day one.

https://lh4.googleusercontent.com/-Q7smhzWnXG0/VK7JKPgwLFI/AAAAAAAAiRM/kQrhYfQAQUA/w997-h665-no/IMG_1671.JPG

https://lh6.googleusercontent.com/-Lx-2MqF0l0U/VK7JKDqGglI/AAAAAAAAiRQ/lRf2Eo2dhbQ/w997-h665-no/IMG_1672.JPG

Hi- I will take mine apart today and see what is going on in there.

Last night I started the two using 250mhs and 800mV  and they have been running at 62.5ghs pretty nicely. about 12 hours later the one ZOMBIED.  I also created even MORE cooling to the heatsinks in like a custom wind tunnel design...

So I will take it apart and clean it. I have very good expeirence in electronics so I wont noob this up. Also got all the correct tools and liquids etc to do a proper job.

I suspect it wont help... the next route after this will be a powered USB hub. And I will connect that hub to the PC without the 5V channel and let the HUB power the units. Also put them on a 4 hour timer. turn of for 15 minutes and restart the whole thing. If I turn of the USB hub power then that should emualte pulling out the cable and restarting the MCU on next power up. I got all this on a sperate power adapter so I will just turn everything off, restart cgminer with some scripts and restart it. Then I can keep these settings at 62.5ghs and it will have some short resting periods (i hope that wont cause to much PCB strees) and it will restart and keep running till the cows come home.

OH- by the way, the same thing happened to me when it droped from ~60 to ~45 because of the voltage settings. Recomended is close to  800mv using a nice PSU like a computer one. I run mine on the stock PSU (240v) using 790mv and it got hot but didnt melt but that worried me so I just found and old PC and pulled the PSU and hooked it up to that. Its more effiecient any way and keep it at 800mv now.

Update:

It's been running almost 24 hours with no zombie but the hashrate is lower now...

https://i.imgur.com/FFzUGOz.jpg

I suspect that the thermal compound is the culprit of the zombie.  It was spread all over the board on top of the surface mounted devices under the heat sink which probably caused the instability issues as it would heat up and create a medium for a short to take place...

Can someone else take theirs apart, clean off like I did, and see if that fixes the zombie issue?

A powered USB hub will likely help, as there is some power draw through the USB port. You're right that windows / firmware / whatever it is controlling the device to the OS / cgminer shouldn't behave like that, but it does. I am not sure which party in that has control over a potential remedy.

USB 2.0 Direct into Motherboard USB Slots, NO Hub, and proper USB shielded USB cable.
U3 12V Powered from 450WATT, separate PSU.

It happens only after some hours and a brief interruption USB should just restart mining, for example. Pull out USB cable, ZOMBIE after 10 seconds, plug in USB cable, ZOMBIE goes away.

In this situation the only way to remove ZOMBIE is to hard power cycle the unit, becuase cgminer does not pickup the USB, or if it does it cannot communicate with it, which means the firmware shat a brick sideways.

If you can turn off USB somehow, then thats the key to hard cycle the U3. I just noticed even with 12volt plugged, the red light inside goes out. Also, I deliberatly sacrificed a USB cable and cut the 5V. I wanted to see if I could just turn the PSU on and OFF which would also power cycle MCU on the U3. But no.. that would be to easy. The MCU is powered by the USB +5 and the ASICS are powered by the 12volt line.

Please note these are Pure deductions by trial and error. I have not fried anything, yet. Phew.

On what do you base that they are not all defective? No one has replied here yet saying that their unit does not ZOMBIE at least once a day.

What type of hub are you on, powered or not powered? USB 3.0 or 2.0? There is also some newer firmware being tested which may fix some of the soft images, but zombie'ing is usually a brief interruption on USB.