Power, Heat, Efficiency And Value
More parts make a hungrier machine, efficiencies of modern architectures aside. A second graphics card adds 152W of power draw to my $1750 machine compared to Don’s $1300 PC, while half as many CPU cores make Paul’s machine appear a true miser.
Bringing efficient architectures back into the spotlight, Paul’s weak GPU pulls similar power compared to Don’s single, fast GeForce GTX 970.
The $1750 PC doesn't live up to the hype of Intel's "Devil's Canyon" thermal optimizations, though the same processor produces dramatically lower temperatures in Don's $1300 PC. Then again, I don't believe Don and Paul measure there CPU and GPU temperatures simultaneously.
A similar phenomena occurs for different reasons in the $1750 PC's upper graphics card, which draws some heat off the back of the machine's lower card. The lower card reached a lower temperature (not shown) that was remarkably similar to Don's single GPU.
Using Paul’s $700 machine as the baseline, we can see just how inefficient its hardware has become by modern standards. Noting that the power chart showed its mid-range graphics card exhibiting the least-competitive power numbers, we’ll credit Don’s single GeForce GTX 970 for his machine’s chart-topping energy-to-performance ratio.
Similarly, I’ll blame my second card for a relative loss of efficiency compared to Don’s. Twice as many cards choke down twice as much energy without delivering two times the performance.
For the first time in recent memory, Don’s mid-priced machine tops our complete system value chart. That’s exclusively due to overclocking, where the $700 PC had a locked CPU. On the other hand, understanding the poor cost-to-benefit ratio of high-end components makes me feel pretty good about my 87% value score.
The price of parts that don’t improve performance has the biggest impact on the cheapest machine. We all had to pay $100 for our operating systems, yet that expense was 14% of Paul’s budget and less than 6% of mine. Removing the cost of Windows drops Don’s overclock to a 99% value score, giving Paul his title back.
A single loss won’t keep Don from reclaiming his title. He paid for an extra storage drive that he didn’t use, and taking that cost out of the equation (along with each machine’s optical drive and case) puts his machine back in the value lead.
Long ago, Don noted that some of our gaming benchmarks were bottlenecked by non-graphics components at super-high frame rates that aren’t even relevant to real-world game play. Back then he suggested a value chart for high-end gamers with only the top resolution shown. Since I blew a bunch of cash on a pair of graphics cards, his suggestion still sounds like a good strategy for me to eke out a “value” win.
I’m not going to overplay my hand on this one because we all know that Don’s $1300 collection of components is the true winner in this System Builder Marathon. It gets there by overclocking, but still fell within a few percent of the cheap machine’s value at baseline frequencies. And, his $1300 build is the one that presents the best balance of gaming, encoding, content creating, compiling and file compression performance.
