Overclocking the Pentium G3258 – The Basics
Two cores, low power consumption, good overclocked performance in single threaded applications – what’s not to like about the Pentium G3258? It’s a fantastic processor in its own right, however, when we take price into consideration, it becomes a no-brainer purchase for a HTPC or casual gaming. The overclockability of these CPUs also makes them fun to experiment with. That brings us to the purpose of today’s musings, a guide on how to overclock the G3258.
The basic principles of overclocking these CPUs are similar to Haswell – at least from a perspective of which voltages to adjust. The main difference being the lower physical core count and lack of hyper threading. While this has an impact on threaded application performance (naturally), it does equate to lower operating temperatures and very low power consumption. Unlike their Haswell brethren, if one is using good forms of cooling, it is possible to run out of voltage headroom before running into thermal limitations.
Expectations/Requirements Rundown
Conservative overclocks of 4.5GHz are achievable on good CPU samples with 1.30Vcore or less with air-cooling. Fully stable frequencies in the region of 4.8GHz are possible with water cooling (closed loop water cooling such as the Corsair H100i). Most of our samples are capable of 4.8GHz with 1.42~1.46Vcore.
Long term effects (such as degradation) of running voltages over 1.40V into these CPUs is not yet known.
We recommend using a good air cooler or closed loop water cooler to overclock these CPUs. Ideally full load temperatures should be kept under 80C (core DTS). Higher temperatures may induce stability.
We use Realbench stress test to evaluate our CPUs for stability. The stress test in Realbench is more difficult to pass than other third party tests such as AIDA. As a result, you may notice higher processor frequencies are possible in other tests. We use Realbench because it is a worst case scenario load and features real-world applications such as Handbrake concurrently with Luxmark while stressing memory to evaluate processor stability. Doing so allows us to gauge frequency expectations more realistically than “lighter load” stress tests.
Power consumption is low. At 4.8GHz power consumption is around 88 Watts under Realbench stress test loads with 32GB of memory at DDR3-2133. Load temperatures at these frequencies are below 80C (DTS) on a good water-cooling setup with 20C ambient temps.
Memory frequencies above DDR3-2133 with 16GB or 32GB of memory can be difficult for some CPUs. VCCSA, VCCIO-D, VCCIO-A and cache voltage can help improve stability for memory. We do not recommend using more than 1.30V for any of these rails long term. In most cases 1.15V should suffice for memory frequencies up to DDR3-2133 and default cache speeds.
For sake of stability and easier setup, memory kits rated for Z87/Z97 are best choice. If plug-and-play operation is desired, sticking closer to DDR3-1600 is wise. Higher speeds may require manual tuning (even with XMP) – as each processor sample is different and there is no magic bullet setting for all samples to ensure stability past stock specifications.
Do not combine multiple memory kits even if from the same model. Use a single memory kit rated at the desired frequency and density.