Of the 6 I just ordered one is running noticeably hotter than the rest. I think I know what I'm doing next weekend.
Temperature difference between the 2 hashing PCB was reported to the engineers and it was reviewed. So far, it is not the hardware caused issues, but rather where and how the miners are being deployed and managed per the engineers.
The temperature sensors are
also rather crude, and on a large scale there will be
plenty of 'miscalibrated' ones. I'd be more suspicious of the sensor reading incorrectly than one board running significantly hotter.
Bitmain wisely chose the LM75a to provide temps to the I2C bus. It's sort of the industry standard, not 'crude' at all:
"The LM75A is an industry-standard digital temperature sensor with an integrated sigma-delta analog-to-digital converter (ADC) and I2C interface. The LM75A provides 9-bit digital temperature readings with an accuracy of ±2°C.
Performance-wise, we should see miners within a few degrees of each other, the 10C variance is extreme. Same air source, same fan, etc.
http://www.ti.com/diagrams/custom_diagram_4_LM75A.gif