Don't think I had done any detailed power measurements for my new system yet, so figured I'd take some measurements and post.
Phenom II X2 550 BE (0.125V undervolt -> idle: 0.85V, load 1.2V)
2x2GB Corsair XMS2 PC2-6400
Sapphire Radeon HD 4670 1GB
Samsung 64GB MLC SSD
Samsung 1TB SpinPoint F1
Idle (XP desktop): 76w
CPU load (Prime95, large FFT): 122w
GPU load (RTHDRIBL @ 1920x1080, 4xAA): 115w
GPU load (x264 decode, 25.6Mbps): 83w
GPU load (FurMark, Xtreme, 1280x1024): 142w
Trying to find something that loads the CPU better than Prime95. Also plan to do some testing with all four cores enabled on the CPU, and maybe test some more real-world loads on the GPU (I think FurMark is supposed to be far, far more demanding than any current games).
edit: Prime95 seems to be about as good as anything else at loading the CPU, so I'm just sticking with that. Anyway, here are some more measurements.
CPU+GPU load (Prime95 large FFT and FurMark): 162w
And here are some numbers with the two extra cores enabled
Phenom II X4 B50 (same clock speed and undervolt I had in the dual core config, only difference is the extra two cores are operational)
CPU load (Prime95, large FFT): 167w
Then I experimented with clocks and voltages a bit to try and get power consumption down to dual core levels. I ended up with the processor clocked to 2.6GHz at a voltage of 1.025V. Full load power consumption for this quad core config was almost the same as the dual core config posted earlier. Seemed to be stable, but I only ran P95 for like 20 mins. Clock speed on each core is lower, but for applications that can fully take advantage of a quad core, performance should be much higher than the dual core at 3.1GHz.
CPU load (Prime95, large FFT): 124w
Might try to see what it can do with all four cores enabled at 0.8V. Maybe 1.6-1.8GHz? Would probably sip power at that clock speed and voltage, even with all the cores.