TheOverclocker Issue 38 | Page 17

your integrated GPU and all dependant features including QuickSync. • AVX (Advanced Vector Extensions) are disabled, so if you have any software that takes specific advantage of this instruction set extension, chances are performance will suffer tremendously. As you can tell, these are not minor issues when put together and for those who are power conscious (which we should all be actually) this may not be an ideal way to run your system for everyday usage. Fortunately, we do have BIOS profiles in all decent motherboards and you may use one profile for gaming and overclocking, and another one for regular media and work, in which case you’re not always dealing with these short comings. TEST PLATFORM Our test platform for benchmarks comprised of a Core i3-6300, 8GiB G.Skill Trident Z DDR4-3,466MHz RAM and an EVGA GeForce GTX 780Ti K|NGP|IN Edition running at its stock GPU and memory speeds of 1,072MHz and 7,000MHz respectively, while the operating system used was Windows 10. For the comparison we ran the i3-6300 at its default frequency with memory at 2133MHZ. We then changed the DRAM frequency to 3,400MHz, then followed that with the CPU at 4.6GHz and DRAM frequency at 3,200MHz. Finally, a 4.7GHz CPU clock was used with a 3,473MHz DRAM frequency. BENCHMARK RESULTS As shown in the graphs, the Core i3 has a lot of untapped performance potential. F1 2015 was a prime example of how bandwidthstarved the CPU is, gaining 23% performance by merely increasing the memory frequency to 3,400MHz. In this instance the gain from increasing the DRAM frequency was larger than that recorded by increasing the core frequency to 4.7GHz. On the other end of the spectrum, Total War: Attila's massive number of units in the game move the bottleneck squarely to the CPU and gains of over 42% were seen by increasing the frequency to 4.7GHz. Dragon Age: Inquisition and GTA V were somewhere in between, showing similar gains where the CPU speed or memory speed had somewhat similar benefits. 3DMark 11, despite its age and its largely synthetic nature, remains a relevant indication of a system's gaming performance. It showed sizeable gains with a bump in memory speed and even larger gains with the processor overclocked. Overall, we were able to get an extra 15% out of 3DMark. Looking at the multi-threaded CPU benchmarks, 3D Particle Movement saw a gain of 10% with the increased memory frequency and a massive 25% with the CPU overclocked as well. CineBench R15 tells a similar tale. With just the memory frequency increased it offered a performance increase of just under 3%, but with the CPU running at 4.7GHz that was extended well over 26%. wPrime 1024M showed absolutely no gain from increasing the memory speed at all, but a 19% gain from increasing the CPU frequency. This puts it in approximately the same territory as a stock Core i5-6600K. SuperPi 32M showed an improvement of almost 5% from a memory overclock alone, which jumped to an improvement of almost 22% when operating at 4.7GHz. It is worth noting that in single-threaded tests and benchmarks, it would keep up with more expensive and higher core count CPUs such as the Core i5-6600 or Core i5-6600K. TAKING IT UP A NOTCH The results we've shown have all been air cooled and at settings safe for 24/7 use for this particular CPU, but that isn't where things end by any measure. The massive Bclk headroom available on Skylake Issue 38 | 2016 The OverClocker 17