GPU overclocking: worth the bother?
You’ve got the processor out the way, so what’s next? Memory aside (we’ll cover that later), it’s all about the graphics card. For gamers, this piece of hardware is crucial when it comes to pumping out frames. Most graphics cards are capable of being overclocked — some are designed for it, others less so. Overclocking a GPU generally increases average frame rates, but more importantly, minimum frame rates, too. So you’ll have a smoother experience and, in the heat of battle, screen tearing won’t become another nemesis.
BACK TO THE BLACKBOARD
Overclocking a GPU shares a lot of similarities with overclocking a CPU. Generally, you should always start at the same point: research. Know what your GPU is capable of, what architecture it’s built from, and how well it overclocks. Unlike CPUs, GPU architectures vary wildly in overclocking potential.
We’re using a reference Nvidia GeForce GTX 1080, with the Pascal architecture, which is very similar to the overclocking king, Maxwell, or the 900 series.
OC PROFILES AND PRE-BINNED CHIPS
Are OC profiles worth it? Well, similar to the automatic versus manual debate earlier, these profiles differ very little. ASUS, MSI, Gigabyte and EVGA set out a specific speed profile for each generation of card ahead of time, and bake it into some form of GPU overclocking software. This can work well, if you’re lucky with the silicon lottery, but can fall flat if the processor doesn’t hold well, and you end up with a card incapable of meeting its slightly overcompensating overclock profile. There’s often no difference between a pre-overclocked GPU and a standard one — you’re probably just applying a more aggressive overclock profile than you need.
Then there’s pre-binning. All chips are binned upon manufacture. Think of those silicon wafers: Each one houses an entire generation of processors, because location and variance within the silicon dictates which chip is what. The middle of the wafer is always better quality than the edge. And this is where the manufacturer discerns which processors will be 1080s, 1080 Tis, or Titans. These are sold to aftermarket partners. Now, these manufacturers can either just place them in the cards that they’re assigned to, or bin them further. To do this, they test which cores leak the most voltage, which chips require the least volts to operate, and which chips overclock the highest. The very best go into the high end ‘premium’ cards, such as EVGA’s FTW cards, while the lesser ones go into more mid-range and reference cards.
COOLING AND AIB PCB
You also need to consider what style of card you want as regards cooling. There are four styles. The standard blower design features a single intake fan, drawing air into an enclosed shroud, pushing it through a heatsink, and out the back of the rear I/O — brilliant for cases with limited space. Then there are the dual- or triple-fan dissipation cards. These are often found on the heavy hitters — ASUS’s DCU III, Gigabyte’s Windforce and MSI’s Twin Frozr are good examples. These cards draw air in and around via the main fans, push it down through a heatsink and out into the surrounding case, and are often quieter and cooler than their blower counterparts.
Lastly, you get the fully water-cooled cards. These either come as a closedloop card, such as the Fury X, or with preinstalled water blocks, such as the EVGA Hydro Copper, requiring you to insert them into a custom water-cooled loop. And on top of all that, there are the miscellaneous hybrids, often utilising a combination of both air and water cooling.
There are aftermarket cards touting more advanced PCB solutions, too. These often include more advanced MOSFETs and chokes, in combination with a better power phase design, meaning better overclocks and higher performance as standard.
“Overclocking a GPU shares a lot of similarities with overclocking a CPU. Generally, you should always start at the same point: research.”