Component Installation And Overclocking
The installation guide of our Phanteks CPU cooler review provides all of the information needed to attach it to an LGA-1150 motherboard. Since I’d already forgone my selected G.Skill Ares memory modules in favor of the slightly cheaper (and better-rated) Ripjaws X, the forward fan was initially installed with its top protruding above the sink.
One wouldn’t expect that to cause any installation problems in an $80 gaming case, but I only paid $56 for this one. And it should cost $56, because that extra space didn’t leave room to close the side panel. Rather than take a hacksaw to the RAM, I reversed the fans.
I learned a long time ago that fans push air towards the center, but draw air from around the sides of the frame in a funnel-shape. That means they don’t draw air through a sink well, unless that sink is properly shrouded. Since this would reduce performance, I reversed the direction of both CPU fans and flipped the case’s exhaust fan around to act as a secondary intake. Two intake fans (one front, one rear) is still a viable (if imperfect) option since this case is also well-ventilated on top.
I also found that two of the Chaser A31’s exhaust fan holes were stripped during factory assembly, which is an expected defect in $56 (but not $80) cases. Reversing the fan gave me access to pristine plastic.
Initial delight that this CPU would easily clock to 4.70 GHz at 1.28V without crashing turned to frustration as the CPU got hot and began to throttle. A cool-to-the-touch heatsink and barely-warm motherboard was number 1 indicator that this CPU core wasn’t transferring heat properly between the core and its so-called “heat spreader”. Frustrated over thermal transfer promises broken by Intel’s “Devil’s Canyon”, I settled for 4.60 GHz at 1.26V.
That also helps answer the “Why so big?” question frequently asked by readers of our cooling articles. A relatively large decrease in surface temperature results in a fairly small decrease in die temperature. As for the idea that the stock cooler must have been good enough for the CPU’s rated 4.40 GHz, Intel Turbo Boost steps down as multiple cores are loaded, and motherboard-provided turbo boost enhancements (full speed with all cores loaded) often run up against thermal throttling in otherwise-stock systems.
My fixed 1.26V core is barely higher than the 1.25V maximum the processor would have produced without intervention. That tells me I should have stuck to “adaptive” mode and “offset” voltage to reach my 4.8 GHz with two cores loaded and 4.6 GHz with four. Lacking the time to sort out a variable overclock, I’ll try to keep this in mind for my next build.
G.Skill’s DDR3-2133 CAS 8 modules overclocked to the same settings as its DDR3-1866 CAS 8 from my previous build. Unable to go any higher at the rated 1.65V, I tried lowering the voltage to reduce heat without altering my best-found frequency and timings, eventually discovering the same 1.60V optimum as used with the previous set. Though I was hoping the new modules would overclock better, this consistency saves configuration time.
Graphics stability tests showed that this GTX 980 spent most of its time power-throttling, and increasing the power threshold provided a huge advantage in performance consistency. The GPU’s eagerness to run 250 MHz over-stock eventually gave way to a crash two hours into Grid 2, so I dropped it to 240 MHz. Similarly, graphics memory ran at the +500 MHz setting for a couple hours before producing artifacts in Battlefield 4, so I dropped it back to +480 MHz. Just as with those artifacts, these frequency settings are showing me patterns where none should exist!