The three different loads are:

Idle - X6800 @ 2.93 GHz, one X1900XTX, Memory at 2.25V - IDLE

Load 2 - X6800 @ 2.93 GHz, one X1900XTX, Memory at 2.25V - LOAD

Load 3 - X6800 @ 3.64 GHz 1.5V, X1900 CrossFire, Memory at 2.25V - LOAD

The idle test is with the system close to stock and just left idling in windows,. This is simply a reference point for all the voltages for the load tests.

Load 2 and 3 are tested with Prime95 running on core 1, and 3D Mark 2005 looped on core 2. The system is left at full load for 30 minutes before taking measurements for each rail with a multi meter.

Results

First up, the FSP FX700-GLN Epsilon. As you would expect, the +3.3V and +5.0V are almost nominal, though the real interest should be on the +12V rails. At idle all the +12V rails hover around 12.29V and lower to 12.20V on load 3.

Voltage regulation is a little bit loose all round. On load 3 the average on the 12V rails is 12.20V which is +1.7%. Still, these are all solid results.

In operation the FSP Epsilon is not exactly quiet, but its easily acceptable to most people. While testing at load 3 the fan noise stayed constant.

The Thermaltake Toughpower 650W was quite a surprise if i were honest. It's +12V rails are extremely tight and showed hardly any voltage dips or peaks from idle to Load 3. I suspected this could be a good unit from the voltage regulation specification, but the results measured here are extremely good.

Voltage regulation is better than the FSP with an average +12V of 12.09 on load 3 (+0.75V). More impressive however is how there was not much change through out the load tests.

This power supply is quiet, you can hear it whirr if you really pay attention, but overall it's very acceptable. It also did not increase its fan speed during testing, even with X1900 CrossFire and a big overclock.