Power Consumption, Thermals, Stability
Power consumption was measured at the wall using a Kill-A-Watt power meter; the numbers given are indicative of the system's total power draw. We measure four specific states:
- Idle: The system is booted and left with no background tasks running for 15 minutes.
- Load: This state models real-world power consumption when the machine is being used for computationally intensive tasks, including 3D gaming, rendering, or data analysis. The applications we use to determine an accurate value for load power can vary depending on the component or components being tested. In this case, we measured power consumption while playing Battlefield: Bad Company 2.
- Peak: Peak power consumption is the maximum amount of power we were able to draw at the wall. Unlike idle and load measurements, peak power should not be treated as a realistic measure of a system's normal power consumption.
We recognize that both ATI and NV have a valid interest in protecting users from melting their video cards when they fail to heed the various all-capital warnings that convey the benchmark's nature. Nevertheless, Furmark is a very useful product for reviewers. It allows us to verify the efficacy of GPU cooling solutions in worst-case scenarios, Equally vital, it allows us to simulate the presence of dust, inadequate ventilation, or fans spinning slower than they should be. Components in such a system can easily run 10-15'C hotter. Toss in an ambient temperature of 75-80'F instead of the 60-65'F we test at, and systems that were stable in the lab when brand new can turn unstable in the field 12-18 months later.
Because we were able to get Furmark 1.9.0 to properly run in SLI mode, we tested as we have before. This may change with future reviews, however.
Idle power, meanwhile, is downright impressive; the Enix draws just 45 percent as much electricity in Idle mode as the Shift or the Genesis.
We were very happy with the Enix's power consumption, but our hopes that we'd see subsequently lower temperatures were dashed. Boutique builders continue to push processors to the absolute limits of their thermal design.
Of the three systems we've discussed here, only the Origin Genesis has been capable of running both Furmark and Prime95 simultaneously for more than eight hours straight. The Shift managed an hour--not a result we were thrilled with, but an acceptable one.
The Enix manages 17 minutes. Even that's only thanks to NV's enhanced hardware throttling that automatically reduced the second GTX 580's workload to ~50 percent of what it was when the test began.
Concerned, we next tested a gaming scenario in which two CPU cores have been left running a task in the background (Prime95 in this case) while the other two cores were occupied playing Left 4 Dead 2. It took the system longer to crash--about 35 minutes—but it did.
What's so bothersome here is that $3000 - $5000 systems carry an implicit guarantee of greater stability, higher-quality components, and hardened design. Indeed they often offer such at stock speeds--but then carve away at the margin of error until it's the thickness of a razor blade. This is not unique to Digital Storm, but we're increasingly unhappy with this tendency. There's no reason a user shouldn't be able to run a program like Prime95 or Folding@Home in the background, while blowing the crap out of zombies simultaneously.