Overclocking Three Partner Boards
Thanks to freely available and easy-to-use tools, overclocking is something of a common pastime among gamers. However, not every tweak makes sense, especially if you want your hardware to enjoy a long and happy life. Since Nvidia's GPU Boost 2.0 is primarily affected by thermals, we want to determine how far we can we can push each card with only superficial software tweaks, while still maintaining a constant boost state using the bundled cooler. We purposely didn’t include the reference card, since it’s already running into its thermal limit at stock settings, and adjusting its fan curve up would have a negative impact on acoustics.
We will note, though, that the reference card maintains an advantage in that its closed shroud pushes waste heat out the I/O bracket. In comparison, all of the partner boards we're testing sacrifice this attribute, leaning on multiple axial fans to blow hot air throughout your case.
EVGA's Precision X is our tool of choice for these overclocking experiments.
Possible Clock Speed Gains
Initially, we tried to determine the maximum frequency boost for each card by testing at stock settings, and then repeating the benchmark with the power target set to 105% and 109%. Although the thermal target should not become an issue, since none of these cards exceeded the 79°C limit (thanks to their coolers), we raised the thermal ceiling to 94°C. Interestingly, the resulting ceilings are virtually identical for all three cards, regardless of their individual stock settings.
On average, we were only able to bump the core clock by another 50 MHz, even with the power target set to 109%. We even tried applying a slight voltage tweak to the MSI card, the disappointing result being a small increase in chip temperatures and the loss of a stable boost clock.
Noting that there’s a clear divergence between the GPU clocks the user can select and the actual boost frequencies that can be sustained, we set our three partner boards to a readily-achievable 1,200 MHz core clock and logged the boost clocks and temperature over the course of two hours.
Although we set all cards to the same speed, the resulting boost GPU Boost rating differs between all three models, with Gigabyte taking the lead. Rethinking our approach, we decided to determine the highest maximum boost instead. Incrementally increasing the settings for each card yielded maximum frequencies between 1,306 and 1,308 MHz before hitting a wall, regardless of the selected GPU speed. Thus, all three cards share a virtually identical upper limit.
The temperatures remain practically unchanged, even when the cards are overclocked. Then again, that’s no real feat, since a mere 50 to 67 MHz is not a huge increase. Meanwhile, performance improves only marginally in our benchmarks. Certainly, it’s not something you would notice in games and hardly justifies an increase in power consumption by 8 to 15 W.
While it may be possible to run the cards at high frequencies for the duration of a benchmark or two, we had to resign ourselves to the fact that such an achievement was only short-lived, with the clock speeds beginning to decrease after 30 to 60 minutes at best. We should have expected this, since the same held true for the GeForce GTX 680. However, all of the results quoted above were verified to be stable over the course of two hours for each card.