The bank balance is reduced to smoldering ruin, you’ve got your new or upgraded rig with top-end components, and now you say you want more? Is it even possible to please you people?

Didn’t think so.

Welcome to the ThinkComputers overclocking guide for the NVIDIA GK104. Also known as the NVIDIA GeForce 680, 670, and 660Ti! This guide does MAY apply to other cards in the series, but, use it at your own risk, as it has not been tested on those devices. If this is your first time overclocking, we hope this guide will be useful to you. If you’re running Intel’s latest processor technology, Ivy Bridge, feel free to check out our sister guide that’ll help you in that area.

Overclocking a video card is a similar sort of affair to overclocking a CPU; there are less variables to play with, but the process is very similar. In this guide I’ll be detailing just how to overclock your sexy new GPU, potentially reaching new heights and beating your friends out in benchmarks with their “stock” configurations! Hah!

Keep reading, and we’ll detail a few methods you can use as well as some instructions on how to go about overclocking your equipment.

Disclaimer: This is an information resource. ThinkComputers cannot and will not be held liable for any damage occurring to your hardware as a result of using this guide. All hardware varies to a certain degree and what others are able to attain, you simply may not be able to. Always be careful, make sure your cooling capabilities can keep up, and don’t overvolt if you don’t know what you’re doing.

MSI Afterburner
An ever popular tool, MSI Afterburner is a little piece of software with a small footprint that can be used for overclocking both NVIDIA and AMD GPU’s. It’s main interface is simple and straightforward, and can be described as my tool of choice when it comes down to graphics card overclocking.

Once installed and opened, MSI Afterburner will present you with sliders for Power Limit, Core Clock, Memory Clock, and Fan Speed. You can leave the fan speed on auto if you want to, or you can give it some more revolutions per minute in order to have your cooler perform better. Yes, cooling affects you overclockability, so make sure you’re adequately cooled.

The first thing you’re going to want to do is unlock voltage tweaking. To do this, click the settings button along the bottom of the softwares GUI. The setting required is under the “General” tab, with the header Compatibility Properties.

This will make voltage control available in the software.

The card being used for this guide has a locked maximum voltage, so I’m able to only add an extra 100mV to the core. You most likely won’t have an issue adding the full 100mV, if yours also has locked voltage, but always be careful and go up in increments. If you’re using a custom BIOS or a card with unlocked voltage like the MSI Lightning series, then you’ll be able to add more. Make sure you are aware of the maximum voltages that you can put through the nVidia chip, as too much is going to cook it. Of course, don’t just jump to the maximum recommended voltage either. You can also increase the TDP, which is referred to as Power Limit. This basically gives you some headroom for power consumption so that the card doesn’t throttle down when it reaches a high power draw.

Right, then! The fun part.

One at a time. Do not increase both memory and core clock at the same time. If you do, it makes troubleshooting any issues a lot more complicated. Get a stable clock on one before you do the other.

I’m going to start with Core clock, but feel free to do this in any order you prefer. Start by dragging the slider to the right to increase the speed of the core. I recommend going in small increments of 25 Mhz or so. Use a GPU-stressing application of your choice to verify stability. I like to use 3DMark11, or, Unigine Heaven (fully maxed out.) Keep going until you get a lockup, a driver crash, or artifacts. Artifacts can be easily described as something that shouldn’t be there. Weird colors flying around the screen, screen tears, etc. Below this section, I’ll leave you a picture of some. Once one of these issues crops up, you’ve pushed the clock too far. Scale it back a bit.

Time for memory! Make sure your memory blocks and VRMs have adequate cooling. Memory that gets too hot can cook. If you don’t have temperature monitors on your memory, I wouldn’t recommend going TOO far, even if it’s stable. Same concept however. Gradually increase your memory clock speed until you get artifacts, lockups, or driver crashes. When that happens, pull it back.

As promised, here’s what artifacts look like: