Overclocking

download msi afterburner and atitool. both are free. next, check the max temp of the card via nvidias site on the specs page. in afterburner, monitor the temperature. in atitool, run the spinning cube. dont click out, or the cube stops. while its spinning, monitor the temps. dont let it get within 10 degrees c of the max temp. if the temperature after 10 minutes of runing the cube shows you have enough of a difference between it and the max temp, increase the core clock slider on afterburner by 10 mhz at a

there are a lot of methods to OC, and the above method isnt bad. ill share my personal method too:

afterburner is a sexy tool, and youll want it due it all it info it logs and what it allows you to change, namely the voltage, core freq, ram freq and shader freq.

first off, research what the max temp is for your specific gpu as specified by the manufacturer. when your GPU in your computer reaches its peak temperature (under full load in gaming on a hot day), youll generally want it at least 10 degrees less than the specified max temp. 20 degrees less is generally much safer. also research what others with your GPU consider a good daily temperature to give you a general idea of whats considered safe.

different games respond to an OC'ed GPU differently; some games will allow a higher core frequency than others, for example. this isnt a very common occurrence by any means, but it will mean your max OC wont always play nicely with all games, which is why i prefer to use real games for my OC testing. after all, youre overclocking your GPU for better gaming performance, right? may as well use playable games to try 'breaking' your OC. (DISCLAIMER: this method is way slower than the atitools or general stress test method, and requires more fine-tuning from you, the user. it does however allow you squeeze out a bit more real world performance).

in the beginning, voltage: more voltage = higher possible frequencies & always higher temps. the only time you raise voltage is if youve 'hit a brick wall' and your then current temps are acceptable; in other words, you raise voltage when you cant increase the core frequency anymore with out getting errors or having something crash, AND you think your GPU would be fine accepting more heat. if you cant raise the core frequency anymore, and your GPU is already running hot, increasing voltage is a bad idea without adding extra cooling (can cut down the components life or just flat out kill it dead).

another thing to consider is the difference between GPU core performance and GPU vram performance. sometimes overclocking both at the same time cuts down the max freq. of each respectively. in other words, sometimes overclocking the GPU core and NOT the vram at all allows the GPU core to reach a higher freq. than if you OC'ed both the core and vram together. it is wise to overclock both the vram and the core, but if your particular GPU is kinda deficient in one area, you should try to bias your OC to balance it out. in other words, find what area of your GPU could do with the most improvement, and tend to focus a little more on that.

another helpful point is to google the difference between vram errors and GPU core errors. if youre starting the push your GPU past where its comfortable, youll notice little errors in games, and knowing what component is producing which error allows you to dial back the specific troublesome component (vram, for example). if this sounds like a pain in the ass, just wait till you see an error (a weird polygon or something) and google it. youll eventually learn which components make which on-screen errors.

so with all that in your pocket, start raising crap (not voltage though yet!). frequencies your GPU really doesnt like will produce errors or will crash very quickly in games. when youre inching slowly toward your GPU's max real-world frequency, games will take some time to produce any error if theyre going to produce errors at all. if you think youve found stable OC, you should play a demanding game for ~3+ hours to really test it. if it produces any errors at all during or after that time, dial the frequency down. it could be that literally 1mhz makes the difference! but now the fun part.

you should always be looking at your GPUs temp just to make sure the peak temp (under full gaming load) doesnt come close to its max allowed temp. afterburner logs this for easy reference. so youve dialed down your gpu that 1mhz and now its perfectly stable. congrats. and your peak temp is fairly low or average. time to up the voltage! and thereby up the frequencies a bit. keep upping them using the same method as before, with this time paying extra attention to the temps, because now youve changed a fundamental aspect of the stock GPU and stock GPU cooling isnt made for specs different than stock settings.

well this is all getting a bit wordy and probably confusing, so if this all seems like too much, ignore this and just use the above previous posters method