blackarrow, on 14 Nov 2014 - 11:39 AM, said:
We have now located this issue: It has been introduced by a part that we have used in our latest shipped batch (shipped between 20 Oct to 13 Nov). All miners affected by this problem have a barcode sticker (picture will be posted here:http://www.blackarro...y_hashboard.jpg. We are currently sending individual emails to all customers that we believe that are affected. If you have not received an email but are concerned that your miner is affected, please open it and check if your hashboard has such sticker. If your miner is affected, please stop using it immediately. If you do not stop using your miner there is a very high possibility that one or more hashboards will burn. We are now testing new hashboards in which we believe that we have solved the problem. We will send new hashboards to all customers as soon as tests are conclusive. Alternatively we can make arrangements to host your miner in our partner's datacenter in China that charges an all-inclusive fee.
We are deeply sorry for the inconvenience.
Technical details: On the latest batch of miners we started to use a different mosfet because of higher efficiency and better sourcing availability. However, to achieve this efficiency this mosfet lacks a diode that protects it from short circuiting the 12V with GND in case there's a wrong message received from the controller. A wrong message can only be caused by the noise in the circuit or by a faulty controller. We have encountered this issue back in May but we have been working with Vishay (the manufacturer) to fix it. Initially, when we found the issue, we wanted to return the chips but Vishay declined to take them back claiming that they are considered safe to be used in consumer electronics and Sony is using them in PS4. Upon applying the changes in our design based on their advice and after testing more than 20 miners we believed that the problem has been solved and started manufacturing miners with these chips. However, it turns out that the problem is still there and in some unknown cases it happens and the mosfet burns, causing a short circuit which heats up the board and sometimes melts the PCIE connector.
They promised they would e-mail all the affected customers immediately, but apparently that's just another broken promise from them.
Assholes.