Guide to Overclocking Your Graphics Card


Sometimes you can to unlock the voltage control, there's an option in MSI settings. But you may need go through a lot more hoops to get the same effect (unofficial overclock/volting, adding some files to MSI's folder, BIOS flashing) and some cards are just voltage locked and there's not much you can do about it. Always be very careful when over-volting since it's easy to burn out the card, this particular value varies wildly and should be looked up on a per-card basis.

I'd recommend some other benchmarks besides Kombustor. b/c it can often show you no artifacts or issues, but when running another game / benchmark you'll get a driver crash or BSOD, or what not. Unigine Heaven, 3dmark, ARR's bench and running a few games that are intensive on your specific card. Crysis 3 is a good example for higher-end cards, I couldn't sustain an 1150/1500 OC on my 7950 which ran just fine through 3dmark, futuremark, unigine, and OCCT. ARR also began oddly crashing or driver failing at odd points until I backed down to 1100/1450 and it's been rock-solid since. So, always test on real games and other benchmarks/software to get a wider array of testing.

My 7950 is a good example of a “great” overclocker, the 7970 fits the bill too but both require voltage to be unlocked (Sapphire/MSI generally unlock their voltages, Vapor-X is reported to be locked). By default my card came at 925Mhz Core, and 1250Mhz Memory. I was able to overclock it to 1100/1450 respectively while keeping my temps around 60-70C under full load in high-intensity games without turning the fan up to 100%. OP's bit about fan curve and tuning is awesome to do, but I'm too lazy and left it on MSI's default which works fine for me, YMMV. My overclock puts me at an 18.9% increase over stock core and 16% on the memory, which is a pretty hefty increase without much additional heat output or any degradation. As far as my understanding is the 7900 series is pretty good at overclocking, but some can of course be rather cruddy. Supposedly 1150 is a common overclock but I wasn't able to hit that.

Since I didn't see it mentioned: test one then the other. Ex: Test Core/Shader but don't move your Memory Clock. Once you hit a wall and cannot move the voltage up any more (due to voltage lock, temp, or unsafe voltages) then set your clocks back to stock and re-test using only Memory clock. Otherwise you introduce too many variables and you'll spend a lot longer troubleshooting for that highest overclock. Test in 5-10mhz steps, until you reach a point of failure, from that point either increase voltage (where applicable) if you can't, then revert to your last setting and run it again to make sure it doesn't artifact or fail. Artifacting is when you see strange colors or shapes or anything really out of the ordinary that doesn't belong in the scene and is a sign the graphics card is being pushed too hard (temps or frequency).

Another thing that was touched on, but should be mentioned again is that you will pull more power from your Power Supply, so make sure you have plenty of headroom before you even start! A PSU that gets too much demand can heat up, cause a fire, small explosion, belch black smoke and damage or fry many other internal components so always, always make sure you check first! It'll also make your room hotter due to higher heat, so that's a concern as well especially with August being one of the hottest months for a lot of places (at least in the States). This can also trip your circuit if you've got other devices running high load, so keep in mind more things can (and to some will) happen when doing this, but it in no way means that OCing is a bad thing.

A Note on Temps:

It doesn't matter the card, don't let your graphics card get to 90C. Some cards can handle it but it's an unsafe temp. The higher the temp the more voltage it has to pull from your PSU/wall since voltage leak increases at higher temperatures and the processors in general operate at a suboptimal level when hotter. When you're approaching such heat you can sometimes get better performance by lowering the clock/voltage slightly to cool it off. And there's always a point towards the upper-limits where you'll need a disproportional amount of voltage to bump up the clocks, it's around there that unless you have good cooling, you should call it quits. 80C for Fermi (400 series, 500 too though they're a refresh) is pretty normal. For newer cards you should strive to stay below 80C for extended durations. Some people may prefer their cards running hotter, but from a realistic standpoint you may be damaging your card and hamstringing your performance later on by running it a little faster presently.

And one last thing about cards. There are two designs (for air-cooled) for graphics cards.

Blower Style:


These cards are best employed in cases with bad airflow as they exhaust outside of the case and help assist in venting the case without creating a much hotter environment. But they are almost always significantly louder and their cooling is inferior to traditional fan design. They are however the most common and generally cheapest version of the card you'll run across. But saving $10-20 (with the intent of OCing) for a hotter card with (generally) less room for overclocking can be a bad idea since with better cooling you can make up the performance of a higher cost card while spending significantly less.

Traditional Style:


These cards tend to come at a premium, varying from $10 all the way up to $50 and generally have an overclock applied so that they're faster out of the box. These cards require better case cooling and if there is bad ventilation or the case has poor intake the fans won't work quite as well as they should. They should still work better than the blower style, but they'll also heat up the inside of the case, causing other potential issues. With good airflow management however these cards are always superior and can often more than make up the cost by a higher overclock.

I've Overclocked graphics cards to the maximum threshold and they still lasted more than enough years to see 3-4 generations go by before I ended up replacing it. If it fails, it's generally due to an already present defect and less of overclocking, unless you ignored the above points. In a way this can make it easier to find defects and therefore makes it a lot easier for you as the consumer to get a replacement before the warranty is up. Running a card like this at a higher rate is bound to expose weaknesses in design or fabrication and as a result it’s going to make them fail sooner if there are any such defects. This is good because it’ll happen more than likely within the warranty period as a result and you’ll be able to RMA without having to essentially pay for the card again or pay a hefty fee for the repair that was their fault to begin with. Otherwise you’re likely to have this issue crop up at some time later far after the warranty is expired and you’ll be out of luck at that point unfortunately.

A generally good idea is to put a graphics card “under water” what this means is to water cool that particular component. Now I know a lot of people are going to look at this and freak out a little, suggesting putting water with an electrical computer component probably seems insane to the average person but hear me out. Water has many times the heat capacity than air, how much easier is it for you to feel a shift in the temperature of water than of the air? Pretty easy right? I thought so. That’s because water can hold more heat and also transfer it much faster than air can as air is an insulator. This makes water an ideal substance for whisking heat away from pesky heat generating components.

How this works is water flows over a set of carved channels inside a closed space that sits atop the card. No fans are any moving parts are on the graphics card or CPU or whatever it is you choose to cool. Instead water flows through piping, either rigid, metal or bendy plastic tubing, your choice. This water is pumped through those channels which absorbs the heat coming from the part incredibly fast and at a very high efficiency, and then that heat goes through a lot of little tubes with small fins attached. This is a radiator and it works just like the one in your card. A fan blows over the radiator cooling down the liquid that runs through the tubes, pushing the accumulating heat energy out of the case or at the very least out and away from the radiator which should be fed by *cold* outside air so the temps are cool as possible. This water then returns to the rest of the computer components in the loop. This cools off the computer components several times better.

A water cooled graphics card, when appropriately cooled is going to be around 30-40C under full load provided you don’t live in an incredibly hot climate as cooling can only cool you down near ambient temps not below them. So if your room is 90 degrees Fahrenheit then that’s the coldest your components can get because they’re not being cooled by anything lower than the ambient air. This is usually given as a delta, which is the difference between one or the other of two types of systems. Such that the delta of a water cooled loop is the difference of the temp of the loop compared to the ambient air outside, usually the aim is between 5 and 10 Celsius, which is a good range to get and manages to get quite close to the original or base temp of the room. This temp is compared to the average 80C that even a good air cooled graphics card is going to be placed under, this leads to less overclocking headroom, a hotter running card which lowers its left expectancy (though as I mentioned above not a dramatic difference) and in general just makes the card work harder.

Remember that the hotter a substance is the worse electrical conductivity it has, which means with a computer component you’re going to have less and less efficiency the hotter it gets, this means the colder a component is the better it works regardless of how much or little you’ve overclocked the thing. Now, it’s entirely plausible and possible therein that you’re not going to see a big difference that this bit of performance gives. However it prevents degradation of the component which is one of the main reasons why cooling is so very important for your hottest components. Notably the graphics card, being embarrassingly parallel in its abilities is mostly limited by its heat output. The designers have to put specific safeties to prevent damage to person and computer alike, as well as preventing somebody from easily melting their brand new 700 dollar component into worthless slag.

Because of this, they limit the card based on its temperature, though TDP can be used too which is a worse detriment really than by temp. At least with temp you can get around it with water cooling which means that your only limitation to overclocking is how fast or how well the card was originally made. This eliminates the massive hurdle of heat being the primary reason for not having a huge overclock and in several instances can easily lead to a dramatic increase in value because you’re able to have a far better card by spending more on water cooling.

Article By: Penguin Writer

Computers | Hardware | Reference | How To

QR Code
QR Code guide_to_overclocking_your_graphics_card (generated for current page)