> [ 194.894937] ib_mthca 0000:03:00.0: SYS_EN DDR error: syn=4, sock=0, sladdr=0, SPD source=DIMM According to the Mellanox documentation, this is indicating a calibration error when trying to initialize the HCA-attached memory. I'm not sure exactly what this means, but bad HCA hardware seems like a pretty good guess. - R.