JGZ: Careful. Please don't go round making these statements without justification. It's certainly true that GCC 3.3 will generally generate faster code, but I've seen nothing (yet) to backup "much faster". Someone's welcome to make some benchmarks or point to references to back this up.
bernie: Yes, there is has been considerable pthreads support added since the last official release. However, this support is in Unixlib, so is also present in the 32-bit 2.95.4 version. The number of potential ports that use threading is actually very small.