Power and Efficiency from a Memory Company: Corsair TX650W Review- Page 3/4
February 26, 2008
By
Thomas Soderstrom
Performance
Can a single system pull over 650 watts? Certainly, but given the efficiency of today's processors it can a difficult level to reach. Yet because of the limited number of cable leads provided in the Corsair TX650W, we were forced to use a single-motherboard test system rather than our usual dual-platform configuration. Even with only one motherboard and a pair of SLI graphics cards, we had to use 4-pin to PCI-Express power adapters to complete the system configuration outlined below.
We had to shoot for an overclocking record on our aging Pentium Extreme Edition 955 simply to raise its power consumption further. Keeping that processor down to "only" 80-85C is a dual-fan TEC-equipped Titan Amanda CPU cooler, which itself is a power hungry monster. Topping off the system are two GeForce 8800GTX graphics cards, overclocked to 622MHz GPU and 1000MHz DRAM (2000MHz data rate) clock speeds, to further assist our high power-consumption efforts.
We realize that some people like to plug devices into a running system, such as removable hard drives or other bay devices, and that some graphics processors are capable of drawing a little more power than ours. To gauge the transitional load effect, we waited until our system reached peak consumption before plugging in four WD Caviar hard drives simultaneously. With a combined start-up power of 50W, we've kept these drives around specifically for this purpose.
Now we had to find the "perfect application" for extracting the highest amount of power from our system. Sniffing around the 3D Mark 2006 test suite revealed that the basic Pixel Shader test, though brief, was otherwise perfect. Fortunately, 3D Mark 2006 also allows repeating a test, either set to a number of times, or run continuously.
But simply forcing the 3D Mark 2006 Pixel Shader Test to loop continuously still wasn't enough for us, as our power supply only consumed around 812W. At 81% efficiency, output power reached *only* the power supply's rated capacity of 650W at default benchmark settings. By choosing a 1920x1200 pixel output resolution with 8x Anti-Aliasing and 8x Anisotropic Filtering, we were able to increase system power consumption to 850W, for a power supply load of 680W.
Adding the transitional-load hard drives during peak system load brought power consumption to 900W, resulting in a 729W power supply load at its rated 81% full-load efficiency. And here are the results.
Rail Output Voltage (dual-platform system)
Rail
Input
+12V
+5V
+3.3V
Idle
548W
12.20
5.06
3.41
Load
850W
12.15
5.05
3.42
Trans.
900W
12.13
5.01
3.39
Even at 850W input power (680W estimated power supply output), the TX650W was very stable with only 0.4% droop on the 12V line. Adding the "50W" transitional load had little effect, except that its 5V draw caused the 5V rail to droop by a tiny 0.7%.
The only thing that remains to be tested is noise level. Our test system was rather loud to begin with (estimated 54db), so how does the TX650W compare? Corsair has another helpful chart.
At 37db full fan-speed, the ambient noise of the TX650W wasn't overly noticeable when compared to the system fans. Buyers with sensitive ears should remember that our system was specifically designed to overstress this power unit, so that the choice of a Core 2 Duo or even a Core 2 Quad, in addition to a single high-end graphics card, would likely bring the noise level down to around 21-26db.