Before detailing my overclocking efforts, first let’s talk about my system's airflow. I chose to mount the power supply so that it took in fresh air through the bottom of the enclosure, rather than acting as an additional exhaust. As a result, Antec's VP-450 was basically removed from the airflow equation. This was fine; I already had a graphics blower and two case fans pushing out air, without any active intakes.
I’m always leery of vacant fan mounts, especially when they're in close proximity to exhaust fans. They very often become unwanted intakes, which impede front to rear airflow. A quick test with tissue paper confirmed my suspicion that the empty 140 mm opening up top was indeed acting as an intake to the adjacent 140 mm exhaust. This probably provided some cool air around the CPU socket, so I chose to test as-is, but folks who don’t procure any optional fans may want to experiment by covering that top vent, forcing more air through the front mesh and side vents.
There isn’t much to say about overclocking the FX-6300. I pursued a conventional multiplier overclock on all cores by disabling the Turbo Core and processor power-saving features in the BIOS. After manually dialing in a stock VID of 1.2625 V to make sure MSI's motherboard didn’t over-volt, and setting the processor fan to run all-out, I found our chip reached 4.0 GHz on all cores without any instability or throttling.
I didn’t explore additional voltage, since monitoring software reported that the socket temperature peaked at 64 Celsius and the package temperature reached 57 degrees (an increase of about five degrees over already-warm stock settings). While we wouldn’t surpass the FX-6300’s maximum Turbo Core frequency in single-threaded workloads, the additional 500 MHz on all cores should really benefit our threaded workloads.
Granted, we could very likely toss $30 more into an aftermarket cooler, raise voltages, and bump the FX-6300 up another 400-500 MHz. But I have to throw in a word of caution: Overclock at your own risk! It can be fun and rewarding, but keep the processor and motherboard components cool.
Maintaining 1.65 V, the Kingston memory was stable at 1866 MT/s, but only at slightly reduced 9-10-10-10-28 timings. This was a purely academic experiment, it turned out, since Sandra 2013 reported no memory bandwidth gains. No attempts were made to boost the CPU-NB frequency; this would have yielded minimal gains, and any additional voltage would have further taxed the little cooling solution. If anything, we'd ferret out higher CPU overclocks if we had more headroom.
Turning attention over to graphics, I fired up EVGA Precision and was amazed to find such an aggressive range for the sliders. Topping out at 115% power, 95 degrees Celsius, +1001 on the GPU, and +2000 on the memory, EVGA clearly doesn't hold back on potential tuning. The ball is truly in your court to determine safe limits.
This GK104 GPU topped out at +160, resulting in boost frequencies of 1267 MHz. Minor artifacts and tearing became visible at 1280 MHz. Precision's OSD (on-screen display) was extremely useful, and clued me in that I had to boost the power to 110% to prevent throttling while playing Far Cry 3.
I called it quits at a stable +620 MHz (3623 in Precision) before reaching the limits of my sample's on-board memory. To assure long-term stability, I dialed back the final overclock to 1254 MHz GPU Boost (+150) and 7204 MT/s (+600) GDDR5. As a nice bonus, EVGA's thermal solution adequately cooled the GPU without having to override its automatic fan settings.