Power, Heat, And Efficiency
A newer, higher-efficiency architecture with fewer CPU cores doesn't help power consumption enough to make up for the fact that a second GeForce GTX 670 uses a lot of juice. Sure, it draws less power in applications able to tax the Core i5-3570K. However, pegging the graphics cards with heavy loads negate any advantage we might have seen from Ivy Bridge.
Although the CPU in today's machine generates less heat, we also equipped it with less cooling. Cost-conscious heat sinks and fans put both systems in the proverbial doghouse when they're overclocked and forced to contend with a demanding workload. We had to keep our room below 35° Celsius to keep them from throttling.
Our performance graph is organized by the average of averages, but that blue line matters more today. Gaming performance gains were our real goal, and we’re only hoping that gains there are enough to offset losses in other benchmarks. We used the original $2000 machine's stock settings as our performance and power baselines.
Heat is wasted power, and it looks like the PC with six cores generates heat far faster than the SLI-equipped gaming box when they're both overclocked. We could have guessed as much by simply by looking at the size of their coolers and the difference between baseline and overclocked temperatures, but our efficiency tests confirm our assumption. We end up seeing today's alternative configuration produce more work per watt, in spite of its thirsty pair of graphics cards.