Graphics Overclocking: Getting The Most From Your GPU

The performance of a graphics card is determined by the clock speeds at which its graphics processor (GPU) and video memory operate. Generally, higher clock speeds mean higher data throughput, translating into better performance and a smoother frame rate. Simply put, a higher frame rate is always better. Sixty fps is considered the optimum. However, this is really only a guideline, and sensitivity to motion differs from one person to the next.

The 60 fps mark is a topic of frequent discussion. The argument goes that movie theatres show their features at 24 fps, which happens to be the same frame rate many HD videos use--and both appear smooth to the eye. Depending on their genre, certain 3D games may require a higher frame rate than others. For example, a real-time strategy game such as Tom Clancy’s EndWar or the Command and Conquer series will usually look fine with as little as 20 fps. First person shooters like Far Cry 2 or Call of Duty are a different matter entirely. If your character is running sideways while simultaneously turning, you may find that 25 fps is too low and will cause stuttering, which is often fatal in this fast-paced genre.

Depending on how sensitive your eyes are, many gamers can spot the difference between 25, 35, and 60 fps in the faster sequences of a given title. Enthusiasts usually aim for an average of 60 fps, and with good reason. This speed offers something of an insurance policy. If things get hectic on-screen and the graphics card has to handle a heavier workload, frame rates can drop below the "playable" mark. If a card can sustain a higher average frame rate, chances are it won’t dip below that point as often as a weaker model might.

The frame rate a graphics card can achieve in a game should not be confused with your monitor’s refresh rate. As long as the vertical signal is synchronized (v-sync), the frame rate has to adapt to the monitor’s refresh rate. Flat-screen monitors usually use a refresh rate of 60 Hz, which means the frame rate is capped at 60 fps. The upside to this is that the frame rate is always in sync with the monitor’s refresh rate, but on the downside, more intricate scenes may be forced down to 30 or even 15 fps (divisors of the monitor’s 60Hz). Disabling v-sync lets the graphics card output other frame rates, such as 23 or 37 fps, as well. 

However, since the frequency at which the 3D scene is being rendered is no longer synchronized with the screen, you can end up with mismatched lines, or “tearing.” If you’ve ever watched the camera pan across a scene in an HD video running at 24 fps on a 60 Hz display, you know what we’re talking about. Power users can overclock their graphics cards to help improve performance, especially at high resolutions, thereby avoiding some of the artifacts of insufficient horsepower.

Maximizing Frame Rate Through Overclocking

In order to overclock your graphics card, you will need a good overclocking tool, the right graphics driver, and a sufficiently powerful CPU to help realize the benefit of better graphics performance. At lower resolutions with reduced quality settings, a brawny graphics card will practically always be held back by the processor. So if you’re still using an older or slower CPU, overclocking your graphics card won’t really help you all that much. Luckily, that can easily be remedied by tweaking the CPU as well, giving it more performance to throw at a game’s AI and physics subsystems.

CPUs tend to offer quite a bit of overclocking headroom, too. For example, a typical Core i7-920 with a stock clock speed of 2.67 GHz, usually has no trouble hitting 3.8 GHz, which constitutes a clock speed increase of a full 42 percent. Things are very different where graphics cards are concerned. On an ATI Radeon HD 4870, overclocking the GPU from its stock speed of 750 MHz up to 820 MHz is a decent achievement, even though that makes it an improvement of merely 10 percent.

When trying to stabilize an overclocked CPU, the first (and usually best) approach is to increase its core voltage though the motherboard’s BIOS. However, we strongly caution anyone against casual changes to a graphics card’s BIOS. The reason is simple: heat. While you can crank up a processor’s speed and core voltage with relative impunity as long as you slap a heavier cooler on it, that’s not really an option with a graphics card. Their reference coolers are usually designed to keep the GPU below a certain threshold temperature. Depending on the specific model, that threshold can be anywhere between 80 and 105 degrees Celsius. Obviously, the only way to compensate for additional heat as a result of overclocking is to increase fan speed, simultaneously making the card louder.

Whether or not that works depends on the cooler and its fan speed control. It may be possible to swap out a card’s cooler for a better aftermarket model. On the other hand, even the dual-slot reference designs employed by most companies these days are very decent. As far as altering the fan’s speed profile, it depends entirely on the card’s manufacturer and your specific model whether you’ll be able to adjust it through the driver, a utility, or the card’s BIOS.

For this exploration, we chose two special models from MSI’s line-up. Both feature improved coolers and very good cooling profiles that react to increased clock speeds. To determine whether there were any differences between ATI cards and Nvidia models, we chose one of each, namely a Radeon HD 4870 and a GeForce GTX 260 (with 216 SPs). We then used drivers with integrated overclocking functionality as well as third-party tools to increase their clock speeds.

Create a new thread in the UK Article comments forum about this subject
This thread is closed for comments
    Your comment
  • Hmmmm... I wonder how easy it would be to mod a CPU cooler to fit a GPU...
  • What this article really shows is that it really is a waste of time over clocking as the performance gains are minimal vs the risk of damaging your card etc.
  • @psiboy. the risk is negligable if you know what you are doing, yes you can burn out a card but in 99 percent of cases this only happens when some idiot tries ramping core clocks up by 200mhz and then plays crysis for two hours.
    and the gains are releative. if you can get an extra ten frames per second out of your card, you could move from a hardly playable 20fps to a smooth 30fps, dpending on the game and setup you have.
    i for one would always like to have my kit running at optimal performance. hence my 3.8ghz phenom and 940mhz core 4890.
  • To be fair, even a modest increase of maybe 30Mhz on the GPU and Memory is a little bit extra performance without having to pay a penny.