Component Installation And Overclocking
I stated on the previous page that Corsair’s Graphite 230T has limited space above the motherboard to install a radiator, but that I felt that the mounting hole offset provided by Corsair might offer enough clearance. My backup plan would have been to use alternative holes to add even more radiator offset. Unfortunately, none of that worked. Was it time to get out my drill and make holes of my own?
The little mark next to the forward mounting screw of the right fan is where it touched a DIMM. Unwilling to leave marks all over the outside of the case, I first asked myself “What Would Donny Do”, and immediately crossed that off my list.
When it comes to nicknames like “Hacksaw”, there can be only one in an organization. So, I initially rounded off the outer corners of two fans using a bench grinder. After all, "Grinder" isn’t such a bad nickname, is it?
A little more test fitment revealed that it would be possible to use the case’s intended mounting holes, if only I removed a little more material. Sticking to the grinder theme, I used a die grinder to finish the work. Newegg sells alternative fans without the outer tabs for builders who have more money than talent.
Corsair outfits the Graphite 230T with screw-free mounting options on all bays, though the height of its 2.5” bays left the MX100 SSD with excess wiggle room. Crucial’s included spacer took up the slack.
Everything else fit with room to spare and unrestricted airflow, in spite of the Graphite 230T’s low height. The case still has a hard time living up to its $80 price though, since it’s made from the same extremely thin and weak steel typically found in products selling for half as much. Corsair also left around an inch of extra space between the fan mounts and side panel hooks that could have been used for additional radiator offset, which would have enabled a completely straightforward installation.
Starting out with Intel’s Core i7-4790K meant starting out with a processor that was already tweaked to higher peak voltage levels than we’d normally expect from a production Haswell-based die, and last time we were only able to get 4.6GHz (fixed frequency) out of this 4.4GHz-rated (variable frequency) part. Of course, that variation has something to do with limits, since the silicon supports higher speeds with greater stability under lightly-threaded loads. That gave me an idea.
Rather than try shoot for a 4.6GHz fixed clock rate, why not mimic Intel’s Turbo Boost technology, but with a 400MHz increase? That would allow the system to run up to 4.8GHz in single-threaded tasks, and still give the system 4.6GHz of “punch” under more taxing loads.
I had my fingers crossed for a good DRAM overclock, but it wasn’t in the cards (or the DIMMs for that matter). A mere bump to 2133MT/s required loose 11-12-11 timings for stability. Extra voltage didn’t really help much, and I just left it at 1.60V in frustration after trying everything to find a little more performance. As I recall, the same frequency and timings were stable at 1.54V.
CPU voltage, on the other hand, was very useful. A bump to 50mV over stock allowed the system to run stable at 4.8GHz fixed with all four cores fully occupied…at least until thermal throttling kicked in. That’s partly because it resulted in a 1.35V CPU core. I dropped to +20mV to compensate, and reverted to 4.6 to 4.8GHz, based on utilization. Our stability tests are extremely harsh, so you might think you can out-overclock me using this system. Our test breaks other people’s overclocks.
As with DRAM, graphics overclocking was an exercise in frustration. I’d spend up to two hours with a “stable” overclock, only to encounter a crash. While both cards initially appeared stable at +200MHz GPU and +300MHz graphics RAM, phantom crashes lead me to drop both settings to +150MHz.