I'm seeing spurious test failures on the PPC64 machine at the GCC compile farm:
Failures happen mostly in longer-running benchmarks (e.g. DaCapo). I haven't kept any statistics but eclipse seems to be particularly affected.
The relevant code fragment is:
Linux gcc1-power7.osuosl.org 3.8.8-202.fc18.ppc64p7 #1 SMP Thu Apr 18 14:11:12 MST 2013 ppc64 ppc64 ppc64 GNU/Linux (64 threads)