Adding RAM doesn't always have an effect on the score. If the latency of the memory is the same, and you're going from 2GB to 4GB, it shouldn't really affect the score at all. I don't think any of Geekbench's exercises run the memory through any intensive tests that utilize over 2GB to begin with. Mostly it hammers the memory with page ins/outs, generally only utilizing up to 1.5GB of the memory. If you went from 512MB to 1GB you would see a large difference, but not from 2GB to 4GB. The small variations you are seeing in the score could be attributed to minor variations you will see every time you run the bench mark. I've seen benchmarks run on the same rig three times in the same week with no changes made and produce three slightly different scores. Anything less than a 1% variation in benchmark scores is negligible at best.